From 2e56b060a2ebc091a99ee9996677991ddccd9b4e Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Wed, 25 Mar 2026 11:56:45 -0700 Subject: [PATCH 01/10] Add agentic AI test info --- .../supported-models-and-frameworks.qmd | 28 +++++++++++++++++++ 1 file changed, 28 insertions(+) diff --git a/site/developer/supported-models-and-frameworks.qmd b/site/developer/supported-models-and-frameworks.qmd index d1b25d81af..589ec17c9e 100644 --- a/site/developer/supported-models-and-frameworks.qmd +++ b/site/developer/supported-models-and-frameworks.qmd @@ -81,6 +81,12 @@ Retrieval-augmented generation pipelines. - Tests: RAGAS integration - Refer to [RAG evaluation](#rag-evaluation) for dataset requirements and available tests + +## Agentic AI +LLM-based agents that use tools and multi-step reasoning. + +- Tests: DeepEval integration +- Refer to [Agentic AI evaluation](#agentic-ai-evaluation) for available tests ::: ## Machine learning models @@ -465,6 +471,28 @@ run_test( - **Context Precision** — Measures relevance of retrieved contexts - **Answer Relevancy** — Assesses if the response addresses the query +## Agentic AI evaluation + +For agentic AI systems, the {{< var validmind.developer >}} integrates with [DeepEval](https://docs.confident-ai.com/) for trace-based evaluation of tool usage and reasoning. + +### Available agentic tests + +- **TaskCompletion** — Assesses whether the agent achieves the requested outcome +- **PlanQuality** — Measures if generated plans are logical, complete, and efficient +- **PlanAdherence** — Evaluates whether the agent follows its plan during execution +- **ToolCorrectness** — Validates that the agent invokes the expected tools +- **ArgumentCorrectness** — Checks if arguments passed to tools are correct + +### Test requirements + +| Test | Requires model | Input columns | +|------|----------------|---------------| +| TaskCompletion | Yes (`predict_fn`) | `input` | +| PlanQuality | Yes (`predict_fn`) | `input` | +| PlanAdherence | Yes (`predict_fn`) | `input` | +| ToolCorrectness | No | `input`, `expected_tools`, `tools_called` | +| ArgumentCorrectness | No | `input`, `tools_called` | + ## Python and dependency compatibility The {{< var validmind.developer >}} requires: From cf93c4c642454739c3cde6befd60bc4f2685e269 Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Wed, 25 Mar 2026 14:48:34 -0700 Subject: [PATCH 02/10] Unrelated link fix --- .../install-and-initialize-validmind-for-r.qmd | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/site/developer/model-documentation/install-and-initialize-validmind-for-r.qmd b/site/developer/model-documentation/install-and-initialize-validmind-for-r.qmd index f595c8166d..3398550a75 100644 --- a/site/developer/model-documentation/install-and-initialize-validmind-for-r.qmd +++ b/site/developer/model-documentation/install-and-initialize-validmind-for-r.qmd @@ -12,12 +12,12 @@ listing: fields: [title, description] grid-columns: 2 contents: - - path: https://github.com/validmind/validmind-library/blob/main/notebooks/quickstart/quickstart_model_documentation.Rmd - title: "Quickstart for model documentation" - description: "End-to-end model documentation workflow in R: load data, preprocess, train a GLM model, and run the full documentation test suite." - - path: https://github.com/validmind/validmind-library/blob/main/notebooks/quickstart/quickstart_model_validation.Rmd - title: "Quickstart for model validation" - description: "End-to-end model validation workflow in R: load data, run data quality tests, train a champion GLM model, and run model evaluation tests." + - path: https://github.com/validmind/validmind-library/tree/main/notebooks/quickstart/ + title: "Quickstart for model documentation" + description: "End-to-end model documentation workflow in R: load data, preprocess, train a GLM model, and run the full documentation test suite." + - path: https://github.com/validmind/validmind-library/tree/main/notebooks/quickstart/ + title: "Quickstart for model validation" + description: "End-to-end model validation workflow in R: load data, run data quality tests, train a champion GLM model, and run model evaluation tests." --- Use the ValidMind R package to document and validate models built in R. The package interfaces with the {{< var validmind.developer >}} through `reticulate`, giving you access to the full Python API from R. From 65659fad91cb56575119411e26a060c10a15fecf Mon Sep 17 00:00:00 2001 From: Beck <164545837+validbeck@users.noreply.github.com> Date: Thu, 26 Mar 2026 09:41:17 -0700 Subject: [PATCH 03/10] docs: Workflows User Action v2 > User Action step + Quarto update anchor fix (#1250) * F+R User Action v2 & anchors * Edits * Proofread * ugh * sdfgsdg * Perma workflow step anchors * Updating old anchors * Other anchors identified --- .../workflows/_conditional-requirements.qmd | 11 ++++++-- site/guide/workflows/_workflow-step-types.qmd | 26 ++++++++++++++----- .../workflow-configuration-examples.qmd | 25 ++++++++++++------ .../2024/2024-dec-06/release-notes.qmd | 2 -- site/releases/_how-to-upgrade.qmd | 2 ++ site/support/troubleshooting.qmd | 2 -- site/training/_training-register.qmd | 1 + .../using-validmind-for-risk-management.qmd | 2 +- .../configuring-ai-workflows.qmd | 10 +++---- 9 files changed, 54 insertions(+), 27 deletions(-) diff --git a/site/guide/workflows/_conditional-requirements.qmd b/site/guide/workflows/_conditional-requirements.qmd index 85429514b0..95257ad888 100644 --- a/site/guide/workflows/_conditional-requirements.qmd +++ b/site/guide/workflows/_conditional-requirements.qmd @@ -6,6 +6,7 @@ SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial --> Conditional requirements are required or optional for the following step types:^[[Workflow step types](/guide/workflows/workflow-step-types.qmd)] ### {{< fa cube >}} Model Stage Change [optional]{.smallercaps .pink} + 1. Under **[when these conditions are met]{.smallcaps}**, set your `AND` and `OR` conditions. 2. Add a single independent rule or a linked condition group. These rules and groups can be nested if desired: @@ -13,7 +14,8 @@ Conditional requirements are required or optional for the following step types:^ - Click **{{< fa plus >}} Add Rule** to add an independent rule. - Click **{{< fa plus-minus >}} Add Group** to add a linked group of rules that all must be true to qualify. -### {{< fa arrow-pointer >}} User Action v2 [optional]{.smallercaps .pink} +### {{< fa arrow-pointer >}} User Action [optional]{.smallercaps .pink} + ::: {.panel-tabset} @@ -54,6 +56,7 @@ Conditional requirements are required or optional for the following step types:^ ### {{< fa users >}} Approval [required]{.smallercaps .pink} + 1. Under **[approval group]{.smallcaps}**, click **{{< fa plus >}} Add Individual Approver** to set an individual approver selected from **User Roles**,^[[Manage roles](/guide/configuration/manage-roles.qmd)] **Model Stakeholder Types**,^[[Manage model stakeholder types](/guide/configuration/manage-model-stakeholder-types.qmd)] or **User** type model inventory fields:^[[Manage model inventory fields](/guide/model-inventory/manage-model-inventory-fields.qmd#inventory-field-types)] @@ -72,10 +75,12 @@ Conditional requirements are required or optional for the following step types:^ ### {{< fa bullhorn >}} Broadcast [required]{.smallercaps .pink} + Select from either **Model Stakeholder Types**^[[Manage model stakeholder types](/guide/configuration/manage-model-stakeholder-types.qmd)] or **User Roles**.^[[Manage roles](/guide/configuration/manage-roles.qmd)] ### {{< fa maximize >}} Condition Branch [required]{.smallercaps .pink} + 1. Under each branch's **[conditions]{.smallcaps}**, set your `AND` and `OR` conditions. 2. Add a single independent rule or a linked condition group. These rules and groups can be nested if desired: @@ -84,6 +89,7 @@ Select from either **Model Stakeholder Types**^[[Manage model stakeholder types] - Click **{{< fa plus-minus >}} Add Group** to add a linked group of rules that all must be true to qualify. ### {{< fa clock >}} Wait [required]{.smallercaps .pink} + Select when you would like to [resume workflow]{.smallcaps}: @@ -113,6 +119,7 @@ c. Enter in a **[wait duration]{.smallcaps}** in minutes, hours, days, or months ### {{< fa cloud-arrow-up >}} HTTP Request [optional]{.smallercaps .pink} + ::: {.panel-tabset} @@ -154,7 +161,7 @@ Configure conditional requirements for the following step types: - {{< fa cube >}} Model Stage Change **[optional]{.smallercaps}** -- {{< fa arrow-pointer >}} User Action v2 **[optional]{.smallercaps}** +- {{< fa arrow-pointer >}} User Action **[optional]{.smallercaps}** - {{< fa users >}} Approval **[required]{.smallercaps}** diff --git a/site/guide/workflows/_workflow-step-types.qmd b/site/guide/workflows/_workflow-step-types.qmd index e2abf4c311..dded8cef05 100644 --- a/site/guide/workflows/_workflow-step-types.qmd +++ b/site/guide/workflows/_workflow-step-types.qmd @@ -5,17 +5,20 @@ SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial --> :::: {.content-visible unless-format="revealjs"} ### {{< fa cube >}} Model Stage Change + - Transitions a model into another stage. - Requires available model stages associated with the workflow.^[[Manage model stages](/guide/workflows/manage-model-stages.qmd)] | Field | Description | |---:|---| -| [when these conditions are met]{.smallcaps} (optional) | Add conditional requirements to qualify for this step.^[ [Conditional requirements](/guide/workflows/conditional-requirements.qmd#model-stage-change-optional)] | +| [when these conditions are met]{.smallcaps} (optional) | Add conditional requirements to qualify for this step.^[ [Conditional requirements](/guide/workflows/conditional-requirements.qmd#model-stage-change)] | | [set model stage to]{.smallcaps} | Select the model stage to transition to. | : **{{< fa cube >}} Model Stage Change** step configuration {.hover tbl-colwidths="[40,60]"} ### {{< fa wifi >}} Workflow State Change + + Transitions the workflow into another workflow state.^[[Workflow states](/guide/workflows/workflow-states.qmd)] | Field | Description | @@ -24,19 +27,21 @@ Transitions the workflow into another workflow state.^[[Workflow states](/guide/ | [progress]{.smallcaps} | Indicate the progress of the workflow at this transition into the next state. | : **{{< fa wifi >}} Workflow State Change** step configuration {.hover tbl-colwidths="[30,70]"} -### {{< fa arrow-pointer >}} User Action v2 +### {{< fa arrow-pointer >}} User Action + Creates a button that performs an action on behalf of the user. | Field | Description | |---:|---| -| [display action button when]{.smallcaps} (optional) | Add conditional requirements for displaying this step.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#user-action-v2-optional)] | +| [display action button when]{.smallcaps} (optional) | Add conditional requirements for displaying this step.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#user-action)] | | [action button label]{.smallcaps} | Text that displays on your action button. | | [requested fields]{.smallcaps} (optional) | Include inventory fields alongside your action.[^step-types-1] | | [user action deadline]{.smallcaps} (optional) | Include a deadline for when users should complete this action by.^[Toggle **Enable deadline** on to configure a deadline.] | -: **{{< fa arrow-pointer >}} User Action v2** step configuration {.hover tbl-colwidths="[40,60]"} +: **{{< fa arrow-pointer >}} User Action** step configuration {.hover tbl-colwidths="[40,60]"} ### {{< fa users >}} Approval + Display a preconfigured message to users in the selected **[approval group]{.smallcaps}**. @@ -51,6 +56,7 @@ Display a preconfigured message to users in the selected **[approval group]{.sma : **{{< fa users >}} Approval** step configuration {.hover tbl-colwidths="[35,65]"} ### {{< fa bullhorn >}} Broadcast + - Send email notifications to users in selected groups. - Select from either **User Roles** or **Model Stakeholder Types**.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#broadcast-required)] @@ -63,6 +69,7 @@ Display a preconfigured message to users in the selected **[approval group]{.sma : **{{< fa bullhorn >}} Broadcast** step configuration {.hover tbl-colwidths="[25,75]"} ### {{< fa maximize >}} Condition Branch + - Allows diverging paths based on conditional requirements. - A `default` branch will be automatically included as a catch-all `else` path to ensure that your conditional logic is complete. @@ -83,6 +90,7 @@ To configure a condition branch: Proceed with linking the Condition Branch with your subsequent steps.^[[Link workflow together](/guide/workflows/configure-workflows.qmd#link-workflow-together)] ### {{< fa clock >}} Wait + - Creates a time condition for displaying next available action.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#wait-required)] - Used to enforce a time delay or a calendar date milestone. @@ -103,10 +111,12 @@ To configure a wait step, select when you would like to [resume workflow]{.small : **{{< fa clock >}} Wait** step configuration {.hover tbl-colwidths="[30,15,45]"} ### {{< fa arrow-right-arrow-left >}} Run Workflow + Triggers another workflow indicated by **[select workflow to run]{.smallcaps}**, allowing you to chain workflows. ### {{< fa circle-nodes >}} Webhook + When a workflow reaches a webhook step, the workflow will pause. Listen for a `POST` request to resume the workflow: @@ -137,8 +147,9 @@ Send your `POST` request to the webhook URL and include these headers with your ::: ### {{< fa cloud-arrow-up >}} HTTP Request + -Sends a HTTP request with optional additional conditions.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#http-request-optional)] +Sends a HTTP request with optional additional conditions.^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd#http-request)] | Field | Description | |---:|---| @@ -154,6 +165,7 @@ Sends a HTTP request with optional additional conditions.^[[Conditional requirem ### {{< fa circle-stop >}} End + - Designates the end of your workflow. - Use this step when all of your other steps have been configured and linked together.^[[Link workflow together](/guide/workflows/configure-workflows.qmd#link-workflow-together)] @@ -162,7 +174,7 @@ Sends a HTTP request with optional additional conditions.^[[Conditional requirem [^step-types-1]: - - Select from the drop-down to add it to your User Action v2 step. + - Select from the drop-down to add it to your User Action step. - [Manage model inventory fields](/guide/model-inventory/manage-model-inventory-fields.qmd) Users must have write permissions for the field in order to enter values. @@ -208,7 +220,7 @@ Transitions a model into another stage. Transitions the workflow into another workflow state. -#### {{< fa arrow-pointer >}} User Action v2 +#### {{< fa arrow-pointer >}} User Action Creates a button that performs an action on behalf of the user. diff --git a/site/guide/workflows/workflow-configuration-examples.qmd b/site/guide/workflows/workflow-configuration-examples.qmd index cea959bb99..63d23581f6 100644 --- a/site/guide/workflows/workflow-configuration-examples.qmd +++ b/site/guide/workflows/workflow-configuration-examples.qmd @@ -62,6 +62,7 @@ For more information, refer to [Integrations examples](/guide/integrations/integ Detail views of the demo workflow step types[^11] used in the example workflows: #### {{< fa cube >}} Model Stage Change + In this example, the model transitions into a stage[^12] called [In Documentation]{.bubble}: @@ -83,6 +84,7 @@ In this example, the model transitions into a stage[^12] called [In Documentatio :::: #### {{< fa wifi >}} Workflow State Change + In this example, the workflow transitions into a state[^13] called [Model In Development]{.bubble} and progresses the workflow completion to 25%: @@ -101,18 +103,19 @@ In this example, the workflow transitions into a state[^13] called [Model In Dev :::: -#### {{< fa arrow-pointer >}} User Action v2 +#### {{< fa arrow-pointer >}} User Action + In this example, the user is presented with an action button labelled **{{< fa arrow-right >}} Submit Documentation For Review** that requires them to include a model export in a [model export]{.smallcaps} attachment type field, then input the date the model was **Submitted for Initial Validation On**.[^14] :::: {.flex .flex-wrap .justify-around} ::: {.w-60-ns} -![Example {{< fa arrow-pointer >}} User Action v2 configuration modal](configure_user-action.png){width="90%" fig-alt="A screenshot showing an example User Action v2 configuration modal" .screenshot group="user-action"} +![Example {{< fa arrow-pointer >}} User Action configuration modal](configure_user-action.png){width="90%" fig-alt="A screenshot showing an example User Action configuration modal" .screenshot group="user-action"} ::: ::: {.w-40-ns} -![Example {{< fa arrow-pointer >}} User Action v2 step](example_user-action.png){fig-alt="A screenshot showing an example User Action v2 step" .screenshot group="user-action"} +![Example {{< fa arrow-pointer >}} User Action step](example_user-action.png){fig-alt="A screenshot showing an example User Action step" .screenshot group="user-action"} ::: @@ -121,11 +124,11 @@ In this example, the user is presented with an action button labelled **{{< fa a :::: {.flex .flex-wrap .justify-around} ::: {.w-50-ns .pr3} -![Example {{< fa arrow-pointer >}} User Action v2 submission panel requesting upload of a model export](example_user-action-submit1.png){fig-alt="A screenshot showing an example User Action v2 submission panel requesting upload of a model export" .screenshot group="user-action"} +![Example {{< fa arrow-pointer >}} User Action submission panel requesting upload of a model export](example_user-action-submit1.png){fig-alt="A screenshot showing an example User Action submission panel requesting upload of a model export" .screenshot group="user-action"} ::: ::: {.w-50-ns} -![Example {{< fa arrow-pointer >}} User Action v2 submission panel requesting input of initial submission date](example_user-action-submit2.png){fig-alt="A screenshot showing an example User Action v2 submission panel requesting input of initial submission date" .screenshot group="user-action"} +![Example {{< fa arrow-pointer >}} User Action submission panel requesting input of initial submission date](example_user-action-submit2.png){fig-alt="A screenshot showing an example User Action submission panel requesting input of initial submission date" .screenshot group="user-action"} ::: @@ -133,6 +136,7 @@ In this example, the user is presented with an action button labelled **{{< fa a #### {{< fa users >}} Approval + In this example, the [peer reviewers]{.smallcaps} model stakeholders[^15] assigned to the model are presented with an approval message and a `Data Sources` field to review: @@ -154,6 +158,7 @@ In this example, the [peer reviewers]{.smallcaps} model stakeholders[^15] assign ![Example {{< fa users >}} Approval message](example_approval-message.png){fig-alt="A screenshot showing an example Approval message" width=90% .screenshot group="approval"} #### {{< fa bullhorn >}} Broadcast + In this example, the [Customer Admin]{.bubble} user role[^16] is sent an email notification after a model is deployed: @@ -175,6 +180,7 @@ In this example, the [Customer Admin]{.bubble} user role[^16] is sent an email n :::: #### {{< fa maximize >}} Condition Branch + In this example, models where the field [genai model]{.smallcaps} is set to true undergo a special [legal review]{.smallcaps} model stakeholder[^17] approval process:[^18] @@ -194,6 +200,7 @@ In this example, models where the field [genai model]{.smallcaps} is set to true :::: #### {{< fa clock >}} Wait + In this example, the next step on the workflow waits until the timestamp indicated in a [deployment scheduled]{.smallcaps} date time field:[^19] @@ -213,6 +220,7 @@ In this example, the next step on the workflow waits until the timestamp indicat :::: #### {{< fa arrow-right-arrow-left >}} Run Workflow + In this example, the current workflow initiates another workflow called `On model validation`: @@ -232,6 +240,7 @@ In this example, the current workflow initiates another workflow called `On mode :::: #### {{< fa circle-stop >}} End + :::: {.flex .flex-wrap .justify-around} @@ -255,7 +264,7 @@ In this example, the workflow is designated to stop after running an additional **Workflow steps used:** - - [{{< fa arrow-pointer >}} User Action v2](#user-action-v2) + - [{{< fa arrow-pointer >}} User Action](#user-action) - [{{< fa cube >}} Model Stage Change](#model-stage-change) - [{{< fa wifi >}} Workflow State Change](#workflow-state-change) - [{{< fa users >}} Approval](#approval) @@ -270,7 +279,7 @@ In this example, the workflow is designated to stop after running an additional **Workflow steps used:** - - [{{< fa arrow-pointer >}} User Action v2](#user-action-v2) + - [{{< fa arrow-pointer >}} User Action](#user-action) - [{{< fa cube >}} Model Stage Change](#model-stage-change) - [{{< fa wifi >}} Workflow State Change](#workflow-state-change) - [{{< fa maximize >}} Condition Branch](#condition-branch) @@ -287,7 +296,7 @@ In this example, the workflow is designated to stop after running an additional **Workflow steps used:** - - [{{< fa arrow-pointer >}} User Action v2](#user-action-v2) + - [{{< fa arrow-pointer >}} User Action](#user-action) - [{{< fa cube >}} Model Stage Change](#model-stage-change) - [{{< fa wifi >}} Workflow State Change](#workflow-state-change) - [{{< fa bullhorn >}} Broadcast](#broadcast) diff --git a/site/releases/2024/2024-dec-06/release-notes.qmd b/site/releases/2024/2024-dec-06/release-notes.qmd index 142a96f72d..464cceed00 100644 --- a/site/releases/2024/2024-dec-06/release-notes.qmd +++ b/site/releases/2024/2024-dec-06/release-notes.qmd @@ -1308,8 +1308,6 @@ We've now introduced a custom 404 page where you can perform a search instead. :::: - - {{< include /releases/_how-to-upgrade.qmd >}} diff --git a/site/releases/_how-to-upgrade.qmd b/site/releases/_how-to-upgrade.qmd index 7aa53e95d4..54235bdb4d 100644 --- a/site/releases/_how-to-upgrade.qmd +++ b/site/releases/_how-to-upgrade.qmd @@ -2,6 +2,8 @@ ## {{< fa arrow-up-from-bracket >}} How to upgrade + + #### {{< var validmind.platform >}} To access the latest version of the {{< var validmind.platform >}},^[[Log in to {{< var vm.product >}}](/guide/access/log-in-to-validmind.qmd)] hard refresh your browser tab: diff --git a/site/support/troubleshooting.qmd b/site/support/troubleshooting.qmd index b420b42c7b..fcc7e31a45 100644 --- a/site/support/troubleshooting.qmd +++ b/site/support/troubleshooting.qmd @@ -84,8 +84,6 @@ This error indicates a Certificate Authority issue. In rare cases it can signal Contact your internal IT team. Provide details about your Python environment and ask for help installing the correct CA bundle. - - {{< include /releases/_how-to-upgrade.qmd >}} ## Additional resources diff --git a/site/training/_training-register.qmd b/site/training/_training-register.qmd index 8a7cf5d5e3..aaae2f4694 100644 --- a/site/training/_training-register.qmd +++ b/site/training/_training-register.qmd @@ -3,6 +3,7 @@ Refer to the LICENSE file in the root of this repository for details. SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial --> ## {{< fa graduation-cap >}} Register + 1. **Get your access credentials**: Email to enquire about access. diff --git a/site/training/administrator-fundamentals/using-validmind-for-risk-management.qmd b/site/training/administrator-fundamentals/using-validmind-for-risk-management.qmd index 8c154c56e8..0110612344 100644 --- a/site/training/administrator-fundamentals/using-validmind-for-risk-management.qmd +++ b/site/training/administrator-fundamentals/using-validmind-for-risk-management.qmd @@ -292,7 +292,7 @@ When you're done, click [{{< fa chevron-right >}}]() to continue. ::: 1. Click on the workflow you'd added earlier, then: - - To add a step, drag and drop a new step onto the canvas. For example: {{< fa arrow-pointer >}} User Action v2 + - To add a step, drag and drop a new step onto the canvas. For example: {{< fa arrow-pointer >}} User Action - Double-click the new step to open up the configuration modal. 2. After you're finished with step configuration, click **Update Step** to apply your changes. 3. After you've configured a step, link your example workflow together: diff --git a/site/training/ai-governance/configuring-ai-workflows.qmd b/site/training/ai-governance/configuring-ai-workflows.qmd index 2c126076cf..db4cd7c4ae 100644 --- a/site/training/ai-governance/configuring-ai-workflows.qmd +++ b/site/training/ai-governance/configuring-ai-workflows.qmd @@ -94,11 +94,11 @@ This example intake workflow demonstrates key AI governance patterns: ::: ::: {.f3} -- **Model Stage Change** — Sets model stage to [In Risk Classification]{.bubble} -- **User Action v2** — Collects risk classification materials with deadline -- **Broadcast** — Notifies stakeholders of pending risk rating -- **Approval** — Routes to Model Validation Governance for review -- **Condition Branch** — Different paths for High Risk, Medium Risk, or Default +- **{{< fa cube >}} Model Stage Change** — Sets model stage to [In Risk Classification]{.bubble} +- **{{< fa arrow-pointer >}} User Action** — Collects risk classification materials with deadline +- **{{< fa bullhorn >}} Broadcast** — Notifies stakeholders of pending risk rating +- **{{< fa users >}} Approval** — Routes to Model Validation Governance for review +- **{{< fa maximize >}} Condition Branch** — Different paths for High Risk, Medium Risk, or Default ::: # Approval workflows {background-image="../assets/digitalspace-colours.svg" .title-accent} From 85a56eea3eff0b1c555171a6e244525fea30f2e1 Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Thu, 26 Mar 2026 13:34:42 -0700 Subject: [PATCH 04/10] Address comments from Anil --- .../supported-models-and-frameworks.qmd | 22 +++++++++---------- 1 file changed, 11 insertions(+), 11 deletions(-) diff --git a/site/developer/supported-models-and-frameworks.qmd b/site/developer/supported-models-and-frameworks.qmd index 589ec17c9e..ab9319d529 100644 --- a/site/developer/supported-models-and-frameworks.qmd +++ b/site/developer/supported-models-and-frameworks.qmd @@ -85,8 +85,8 @@ Retrieval-augmented generation pipelines. ## Agentic AI LLM-based agents that use tools and multi-step reasoning. -- Tests: DeepEval integration -- Refer to [Agentic AI evaluation](#agentic-ai-evaluation) for available tests +- Scorers: DeepEval integration +- Refer to [Agentic AI evaluation](#agentic-ai-evaluation) for available scorers ::: ## Machine learning models @@ -475,7 +475,7 @@ run_test( For agentic AI systems, the {{< var validmind.developer >}} integrates with [DeepEval](https://docs.confident-ai.com/) for trace-based evaluation of tool usage and reasoning. -### Available agentic tests +### Available agentic scorers - **TaskCompletion** — Assesses whether the agent achieves the requested outcome - **PlanQuality** — Measures if generated plans are logical, complete, and efficient @@ -483,15 +483,15 @@ For agentic AI systems, the {{< var validmind.developer >}} integrates with [Dee - **ToolCorrectness** — Validates that the agent invokes the expected tools - **ArgumentCorrectness** — Checks if arguments passed to tools are correct -### Test requirements +### Scorer requirements -| Test | Requires model | Input columns | -|------|----------------|---------------| -| TaskCompletion | Yes (`predict_fn`) | `input` | -| PlanQuality | Yes (`predict_fn`) | `input` | -| PlanAdherence | Yes (`predict_fn`) | `input` | -| ToolCorrectness | No | `input`, `expected_tools`, `tools_called` | -| ArgumentCorrectness | No | `input`, `tools_called` | +| Scorer | Requires model | Parameters | +|--------|----------------|------------| +| TaskCompletion | Yes (`predict_fn`) | `input_column` | +| PlanQuality | Yes (`predict_fn`) | `input_column` | +| PlanAdherence | Yes (`predict_fn`) | `input_column` | +| ToolCorrectness | No | `input_column`, `expected_tools_called_column`, `actual_tools_called_column` | +| ArgumentCorrectness | No | `input_column`, `actual_tools_called_column` | ## Python and dependency compatibility From b75399e097a5edd1fd34ba1ac72790007505ef0d Mon Sep 17 00:00:00 2001 From: Beck <164545837+validbeck@users.noreply.github.com> Date: Fri, 27 Mar 2026 10:29:56 -0700 Subject: [PATCH 05/10] docs: Workflow versioning (#1252) * WIP * save point * Edit: Configure workflows * Save point * Edit: Manage workflows * proofread --- site/guide/workflows/_initiate-workflows.qmd | 31 +---------- site/guide/workflows/configure-workflows.qmd | 32 ++++++++---- site/guide/workflows/manage-workflows.qmd | 55 +++++++++++++++++++- 3 files changed, 77 insertions(+), 41 deletions(-) diff --git a/site/guide/workflows/_initiate-workflows.qmd b/site/guide/workflows/_initiate-workflows.qmd index 2561b9d105..5053cdfbf0 100644 --- a/site/guide/workflows/_initiate-workflows.qmd +++ b/site/guide/workflows/_initiate-workflows.qmd @@ -7,7 +7,7 @@ SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial --> By default, the [{{< fa hand >}} Customer Admin]{.bubble} role has sufficient permissions to manually start workflows. ::: -To manually start workflows[^initiate-workflows]: +To initiate workflows manually:[^initiate-workflows] ::: {.panel-tabset} @@ -43,35 +43,6 @@ To manually start workflows[^initiate-workflows]: ::: -### Resume workflows - -To manually resume a workflow in a {{< fa clock >}} Wait state:^[[Conditional requirements](/guide/workflows/conditional-requirements.qmd)] - -::: {.panel-tabset} - -#### On models - -1. In the left sidebar, click **{{< fa cubes >}} Inventory**. - -1. Select a model or find your model by applying a filter or searching for it.^[[Working with the model inventory](/guide/model-inventory/working-with-model-inventory.qmd#search-filter-and-sort-models)] - -1. On the landing page of your model, locate the [active workflows]{.smallcaps} section. - -1. For the workflow you'd like to manually advance, click **{{< fa arrow-right-arrow-left >}} Resume Workflow** to open up the workflow's step overview. - -1. On the **{{< fa clock >}} Wait** step, double-click on **{{< fa arrow-right-arrow-left >}}** to advance the workflow immediately. - -#### On artifacts - -1. On the details page of your artifact,^[[View and filter artifacts](/guide/model-validation/view-filter-artifacts.qmd#view-artifacts)] locate the [active workflows]{.smallcaps} section. - -1. For the workflow you'd like to manually advance, click **{{< fa arrow-right-arrow-left >}} Resume Workflow** to open up the workflow's step overview. - -1. On the **{{< fa clock >}} Wait** step, double-click on **{{< fa arrow-right-arrow-left >}}** to advance the workflow immediately. - -::: - - [^initiate-workflows]: diff --git a/site/guide/workflows/configure-workflows.qmd b/site/guide/workflows/configure-workflows.qmd index f69884b85d..f5f0f325a6 100644 --- a/site/guide/workflows/configure-workflows.qmd +++ b/site/guide/workflows/configure-workflows.qmd @@ -70,6 +70,13 @@ To create a copy of an existing workflow: ## Edit existing workflows +::: {.callout title="Saving edits to a workflow already executed on models or artifacts will create a new version of the workflow."} + +- Workflows executed on models or artifacts will not be updated to the new version. To apply the latest version of a workflow, abort the existing version of the workflow and manually initiate the workflow again.[^6] +- Older versions of a workflow are read-only and cannot be edited or initiated on models or artifacts. + +::: + You can edit a workflow in several ways: ::: {.panel-tabset} @@ -86,27 +93,30 @@ You can edit a workflow in several ways: 1. In the left sidebar, click **{{< fa cubes >}} Inventory**. -2. Select a model or find your model by applying a filter or searching for it.[^6] +2. Select a model or find your model by applying a filter or searching for it.[^7] 3. On the landing page of your model, locate the [active workflows]{.smallcaps} section. -4. Click on the name of the workflow you'd like to abort to open that specific workflow's details. +4. Click on the name of the workflow you'd like to edit to open that specific workflow's details. 5. On the workflow's detail modal, click on the **{{< fa ellipsis-vertical >}}** in the top-right hand corner and select **{{< fa pencil >}} Edit Workflow**. #### Via workflows applied to artifacts -1. On the details page of your artifact,[^7] locate the [active workflows]{.smallcaps} section. +1. On the details page of your artifact,[^8] locate the [active workflows]{.smallcaps} section. -2. Click on the name of the workflow you'd like to abort to open that specific workflow's details. +2. Click on the name of the workflow you'd like to edit to open that specific workflow's details. 3. On the workflow's detail modal, click on the **{{< fa ellipsis-vertical >}}** in the top-right hand corner and select **{{< fa pencil >}} Edit Workflow**. ::: -a. Make your desired changes to step configuration[^8] and step relationships[^9] on your workflow. +a. Make your desired changes to step configuration[^9] and step relationships[^10] on your workflow. + +b. When you are finished, click **Save New Version** to apply your changes. + +c. Enter in your **[version notes]{.smallcaps}** to describe your changes. -b. When you are finished, click **Save Workflow** to apply your changes. ### Delete workflow steps @@ -167,11 +177,13 @@ Deleting workflow steps on workflows active on models may result in malfunctioni [^5]: If a cloned workflow's [workflow start]{.smallcaps} trigger conflicts with an existing workflow, the cloned workflow will be set to manual initiation instead. -[^6]: [Working with the model inventory](/guide/model-inventory/working-with-model-inventory.qmd#search-filter-and-sort-models) +[^6]: [Manage workflows](/guide/workflows/manage-workflows.qmd#update-workflow-versions) + +[^7]: [Working with the model inventory](/guide/model-inventory/working-with-model-inventory.qmd#search-filter-and-sort-models) -[^7]: [View and filter artifacts](/guide/model-validation/view-filter-artifacts.qmd#view-artifacts) +[^8]: [View and filter artifacts](/guide/model-validation/view-filter-artifacts.qmd#view-artifacts) -[^8]: [Configure workflow steps](#configure-workflow-steps) +[^9]: [Configure workflow steps](#configure-workflow-steps) -[^9]: [Link workflow together](#link-workflow-together) +[^10]: [Link workflow together](#link-workflow-together) diff --git a/site/guide/workflows/manage-workflows.qmd b/site/guide/workflows/manage-workflows.qmd index 1ae776e73a..372a02f986 100644 --- a/site/guide/workflows/manage-workflows.qmd +++ b/site/guide/workflows/manage-workflows.qmd @@ -35,6 +35,47 @@ Review workflows associated with your models, and initiate or abort workflows co {{< include /guide/workflows/_initiate-workflows.qmd >}} +### Resume workflows + +To manually resume a workflow in a {{< fa clock >}} Wait state:[^4] + +::: {.panel-tabset} + +#### On models + +1. In the left sidebar, click **{{< fa cubes >}} Inventory**. + +1. Select a model or find your model by applying a filter or searching for it.[^5] + +1. On the landing page of your model, locate the [active workflows]{.smallcaps} section. + +1. For the workflow you'd like to manually advance, click **{{< fa arrow-right-arrow-left >}} Resume Workflow** to open up the workflow's step overview. + +1. On the **{{< fa clock >}} Wait** step, double-click on **{{< fa arrow-right-arrow-left >}}** to advance the workflow immediately. + +#### On artifacts + +1. On the details page of your artifact,[^6] locate the [active workflows]{.smallcaps} section. + +1. For the workflow you'd like to manually advance, click **{{< fa arrow-right-arrow-left >}} Resume Workflow** to open up the workflow's step overview. + +1. On the **{{< fa clock >}} Wait** step, double-click on **{{< fa arrow-right-arrow-left >}}** to advance the workflow immediately. + +:::: + +## Update workflow versions + +::: {.callout-important} +Applying the latest version of a workflow will reset the workflow to the beginning. + +::: + +If you have updated a workflow's configuration and want to apply the newest version of the workflow to your models or artifacts:[^7] + +1. Abort the existing execution of the workflow.[^8] + +2. Manually initiate the workflow again.[^9] + ## Abort workflows {{< include /guide/workflows/_abort-workflows.qmd >}} @@ -46,4 +87,16 @@ Review workflows associated with your models, and initiate or abort workflows co [^2]: [Configure workflows](configure-workflows.qmd) -[^3]: [Manage permissions](/guide/configuration/manage-permissions.qmd) \ No newline at end of file +[^3]: [Manage permissions](/guide/configuration/manage-permissions.qmd) + +[^4]: [Conditional requirements](conditional-requirements.qmd#wait) + +[^5]: [Working with the model inventory](/guide/model-inventory/working-with-model-inventory.qmd#search-filter-and-sort-models) + +[^6]: [View and filter artifacts](/guide/model-validation/view-filter-artifacts.qmd#view-artifacts) + +[^7]: [Configure workflows](configure-workflows.qmd#edit-existing-workflows) + +[^8]: [Abort workflows](/guide/workflows/manage-workflows.qmd#abort-workflows) + +[^9]: [Initiate workflows](/guide/workflows/manage-workflows.qmd#initiate-workflows) \ No newline at end of file From 14b9b599480bcc53b658e64b74e7ce4208af0bac Mon Sep 17 00:00:00 2001 From: Beck <164545837+validbeck@users.noreply.github.com> Date: Mon, 30 Mar 2026 13:17:04 -0700 Subject: [PATCH 06/10] Wording adjust (#1258) --- site/guide/model-validation/manage-artifact-types.qmd | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/site/guide/model-validation/manage-artifact-types.qmd b/site/guide/model-validation/manage-artifact-types.qmd index 0f3b7d35a6..8f3669d081 100644 --- a/site/guide/model-validation/manage-artifact-types.qmd +++ b/site/guide/model-validation/manage-artifact-types.qmd @@ -13,7 +13,7 @@ aliases: #### Artifact types -By default, the {{< var validmind.platform >}} provides three native artifact types: +By default, the {{< var validmind.platform >}} provides three artifact types: Validation Issue : Problems discovered during model validation that reflect errors, inconsistencies, or gaps in data, methodology, implementation, or documentation that need to be addressed before the model is considered reliable. @@ -75,7 +75,7 @@ Custom artifact types support the same features as default types, including stat ## Archive and delete artifact types ::: {.callout-important title="Deletion of artifact types is permanent."} -- Native (default) artifact types cannot be deleted, only archived.[^4] When archiving a native artifact type, existing artifacts of that type will be hidden and no new artifacts of that type can be created. +- Default artifact types cannot be deleted, only archived.[^4] When archiving a default artifact type, existing artifacts of that type will be hidden and no new artifacts of that type can be created. - When archiving a custom artifact type, existing artifacts of that type will remain viewable, but no new artifacts of that type can be created. ::: From 74ab013135cfd42db2c1ea44f4e42010e1f6510e Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Tue, 31 Mar 2026 14:23:28 -0700 Subject: [PATCH 07/10] Documentation for ValidMind MCP (#1221) * Initial draft * Updates after chatting with Hunter * Edits, remove extraneous file embed * Edits after testing * More edits after testing * Minor edits * Address review comment from Hunter --- site/about/glossary/mcp/_mcp-server.qmd | 6 + site/about/glossary/mcp/_mcp-tool.qmd | 6 + site/about/glossary/mcp/_mcp.qmd | 6 + site/guide/_sidebar.yaml | 3 +- .../integrations/managing-integrations.qmd | 28 ++- .../mcp/connect-ai-assistants-via-mcp.qmd | 198 ++++++++++++++++++ 6 files changed, 245 insertions(+), 2 deletions(-) create mode 100644 site/about/glossary/mcp/_mcp-server.qmd create mode 100644 site/about/glossary/mcp/_mcp-tool.qmd create mode 100644 site/about/glossary/mcp/_mcp.qmd create mode 100644 site/guide/mcp/connect-ai-assistants-via-mcp.qmd diff --git a/site/about/glossary/mcp/_mcp-server.qmd b/site/about/glossary/mcp/_mcp-server.qmd new file mode 100644 index 0000000000..d758930933 --- /dev/null +++ b/site/about/glossary/mcp/_mcp-server.qmd @@ -0,0 +1,6 @@ + + +MCP server +: A service that exposes tools and resources to AI assistants via the MCP protocol. The {{< var vm.product >}} MCP server provides access to model inventory operations. diff --git a/site/about/glossary/mcp/_mcp-tool.qmd b/site/about/glossary/mcp/_mcp-tool.qmd new file mode 100644 index 0000000000..739c6cd6e5 --- /dev/null +++ b/site/about/glossary/mcp/_mcp-tool.qmd @@ -0,0 +1,6 @@ + + +MCP tool +: An operation exposed by an MCP server that AI assistants can invoke to perform actions, such as querying models or updating artifacts. diff --git a/site/about/glossary/mcp/_mcp.qmd b/site/about/glossary/mcp/_mcp.qmd new file mode 100644 index 0000000000..0dca829df8 --- /dev/null +++ b/site/about/glossary/mcp/_mcp.qmd @@ -0,0 +1,6 @@ + + +MCP (Model Context Protocol) +: An open standard that enables AI assistants to connect to external data sources and tools through a unified protocol, allowing natural language interactions with structured systems. diff --git a/site/guide/_sidebar.yaml b/site/guide/_sidebar.yaml index b5d7d6dcda..fd3b304124 100644 --- a/site/guide/_sidebar.yaml +++ b/site/guide/_sidebar.yaml @@ -42,9 +42,10 @@ website: contents: - guide/integrations/manage-secrets.qmd - guide/integrations/configure-connections.qmd - - guide/integrations/configure-analytics-exports.qmd - guide/integrations/implement-custom-integrations.qmd - guide/integrations/link-external-models.qmd + - guide/mcp/connect-ai-assistants-via-mcp.qmd + - guide/integrations/configure-analytics-exports.qmd - guide/integrations/integrations-examples.qmd - text: "---" - text: "Workflows" diff --git a/site/guide/integrations/managing-integrations.qmd b/site/guide/integrations/managing-integrations.qmd index 2e63178cac..db6c7c1e8a 100644 --- a/site/guide/integrations/managing-integrations.qmd +++ b/site/guide/integrations/managing-integrations.qmd @@ -33,6 +33,7 @@ graph TD E[Public REST API] F[Testing & documentation] G[Service integration] + H[AI assistants] A <--> B A <--> C @@ -40,6 +41,7 @@ graph TD A <--> E A <--> F A <--> G + A <--> H ``` @@ -194,6 +196,29 @@ Link to tickets in external systems and keep them synchronized. Attach tickets t :::: +--- + +### AI assistants + +Connect AI assistants to {{< var vm.product >}} using the Model Context Protocol (MCP). Query and manage your model inventory through natural language conversations.[^7] + +**Supported** + +:::: {.flex .flex-wrap .justify-around} + +::: {.w-33-ns} +- Claude Desktop +::: + +::: {.w-33-ns} +- Cursor IDE +::: + +::: {.w-33-ns} +  +::: + +:::: ## Key concepts @@ -233,4 +258,5 @@ Link to tickets in external systems and keep them synchronized. Attach tickets t [^3]: [Introduction to workflows](/guide/workflows/introduction-to-workflows.qmd) [^4]: [Public REST API](/reference/validmind-rest-api-vm.qmd) [^5]: [Testing & documentation](/developer/validmind-library.qmd) -[^6]: [Configure analytics exports](/guide/integrations/configure-analytics-exports.qmd) \ No newline at end of file +[^6]: [Configure analytics exports](/guide/integrations/configure-analytics-exports.qmd) +[^7]: [Connect AI assistants](/guide/mcp/connect-ai-assistants-via-mcp.qmd) \ No newline at end of file diff --git a/site/guide/mcp/connect-ai-assistants-via-mcp.qmd b/site/guide/mcp/connect-ai-assistants-via-mcp.qmd new file mode 100644 index 0000000000..a2e66f0717 --- /dev/null +++ b/site/guide/mcp/connect-ai-assistants-via-mcp.qmd @@ -0,0 +1,198 @@ +--- +# Copyright © 2023-2026 ValidMind Inc. All rights reserved. +# Refer to the LICENSE file in the root of this repository for details. +# SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial +title: "Connect AI assistants via MCP" +date: last-modified +--- + +The {{< var vm.product >}} MCP (Model Context Protocol) server enables AI assistants to query and manage your model inventory through natural language, providing seamless access to models, artifacts, templates, and custom fields. + + +::: {.attn} + +## Prerequisites + +- [x] {{< var link.login >}} +- [x] You have your API key and secret.[^1] +- [x] You have a compatible AI assistant installed: + - Cursor IDE[^2] + - Claude Code[^3] + +::: + +## Key concepts + +{{< include /about/glossary/mcp/_mcp.qmd >}} + +{{< include /about/glossary/mcp/_mcp-server.qmd >}} + +{{< include /about/glossary/mcp/_mcp-tool.qmd >}} + +### How does {{< var vm.product >}} MCP work? + +```{mermaid} +graph LR + subgraph ai [AI assistants] + Claude[Claude Code] + Cursor[Cursor IDE] + end + + subgraph mcp [MCP layer] + MCPServer[ValidMind MCP Server] + end + + subgraph vm [ValidMind Platform] + Models[Models] + Artifacts[Artifacts] + Templates[Templates] + CustomFields[Custom fields] + end + + Claude --> MCPServer + Cursor --> MCPServer + MCPServer --> Models + MCPServer --> Artifacts + MCPServer --> Templates + MCPServer --> CustomFields +``` + +Your AI assistant sends natural language queries through the MCP protocol. The {{< var vm.product >}} MCP server translates these into API calls, authenticates using your API key, and returns data from the {{< var validmind.platform >}}. + +### What can I do with {{< var vm.product >}} MCP? + +The MCP server exposes tools for working with your model inventory: + +:::: {.flex .flex-wrap .justify-around} + +::: {.w-50-ns .pr3} + +**Model and artifact operations** + +- List and get models +- List and get artifacts +- Filter by risk level, deployment region, or ownership + +**Custom field operations** + +- List custom fields for models and artifacts +- Update custom field values + +::: + +::: {.w-50-ns .pl3} + +**Template operations** + +- List available templates +- Get template details +- Validate templates +- Duplicate and update templates + +::: + +:::: + +**Example queries you can ask:** + +- "Show me all models that are high risk" +- "Find all the models I am an owner for" +- "List my validation artifacts" +- "What templates are available for credit risk models?" + +## Configure your AI assistant + +::: {.panel-tabset} + +### Cursor IDE + +1. In Cursor, open **Settings** > **Cursor Settings** and navigate to the **MCP** section. + +2. Click **Add new global MCP server** to open your `~/.cursor/mcp.json` file. + +3. Add the {{< var vm.product >}} MCP server configuration: + + ```json + { + "mcpServers": { + "validmind": { + "url": "https://api.prod.validmind.ai/mcp", + "headers": { + "x-api-key": "YOUR_API_KEY", + "x-api-secret": "YOUR_API_SECRET" + } + } + } + } + ``` + +4. Replace `YOUR_API_KEY` and `YOUR_API_SECRET` with your ValidMind credentials.[^1] + +5. Save the file and reload Cursor with the **Command/Ctrl + Shift + P** > **Developer: Reload Window** shortcut. + +6. Return to **Cursor Settings > MCP** and verify that the ValidMind server appears. If it shows as disabled, click the toggle to enable it. + +7. Try asking Cursor questions like: + + - "What models do I own?" + - "Show me all tier 1 models" + - "Register a new model called Customer Churn" + +### Claude Code + +1. In Claude Code, add the {{< var vm.product >}} MCP server (this updates your `~/.claude.json` file): + + ```bash + claude mcp add --transport http validmind https://api.prod.validmind.ai/mcp \ + --header "x-api-key: YOUR_API_KEY" \ + --header "x-api-secret: YOUR_API_SECRET" + ``` + +2. Replace `YOUR_API_KEY` and `YOUR_API_SECRET` with your ValidMind credentials.[^1] + +3. Verify the connection: + + ```bash + claude mcp list + ``` + + You should see `validmind` with status `✓ Connected`. + +4. Start Claude Code and try asking questions like: + + - "What models do I own?" + - "Show me all tier 1 models" + - "Register a new model called Customer Churn" + +::: + +## Troubleshooting + +::: {.panel-tabset} + +### Connection refused or timeout errors + +- Verify your network can reach `api.prod.validmind.ai`. +- Check if your organization uses a firewall or proxy that blocks MCP connections. +- Ensure you are using the correct MCP URL for your environment. + +### Authentication errors (401 or 403) + +- Confirm your API key and secret are correct. +- Check that your API credentials have not expired. +- Verify your account has the necessary permissions to access the requested resources. + +### MCP server not appearing in Cursor + +- Ensure the configuration file is saved at `~/.cursor/mcp.json`. +- Validate the JSON syntax in your configuration file. +- Reload Cursor after making configuration changes. +- Check **Cursor Settings > MCP** to verify the server appears and is enabled. You may need to manually toggle the server on after adding it to the configuration. + +::: + + + +[^1]: [Get your API key and secret](/reference/validmind-rest-api-vm.qmd#get-your-api-key-and-secret) +[^2]: [Cursor](https://cursor.com) +[^3]: [Claude Code](https://docs.anthropic.com/en/docs/claude-code) From 560ac76da4f2d8546f81f640f6a7f0e239342a0d Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Tue, 31 Mar 2026 15:46:19 -0700 Subject: [PATCH 08/10] =?UTF-8?q?Preview=20for=C2=A0filter=20by=20category?= =?UTF-8?q?=20LUA=20filter=20alternative=20+=20add=20category-filter=20ext?= =?UTF-8?q?ension=20to=20repo=20(#1255)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * Preview for LUA filter alternative * Add category-filter Lua extension for release pages This extension reads listing-filter config from YAML frontmatter and injects it as a data attribute for the JavaScript filter to consume. * Add .quartoignore * Apply suggestion from @nrichers * Wildcard paths for .Rmd .quartoignore --- site/.quartoignore | 1 + .../category-filter/_extension.yml | 6 ++ .../category-filter/category-filter.lua | 59 +++++++++++++++++++ 3 files changed, 66 insertions(+) create mode 100644 site/.quartoignore create mode 100644 site/_extensions/category-filter/_extension.yml create mode 100644 site/_extensions/category-filter/category-filter.lua diff --git a/site/.quartoignore b/site/.quartoignore new file mode 100644 index 0000000000..cdef88f3fe --- /dev/null +++ b/site/.quartoignore @@ -0,0 +1 @@ +**/*.Rmd \ No newline at end of file diff --git a/site/_extensions/category-filter/_extension.yml b/site/_extensions/category-filter/_extension.yml new file mode 100644 index 0000000000..f12627f1d7 --- /dev/null +++ b/site/_extensions/category-filter/_extension.yml @@ -0,0 +1,6 @@ +title: Category Filter +author: Nik Richers +version: 1.0.0 +contributes: + filters: + - category-filter.lua diff --git a/site/_extensions/category-filter/category-filter.lua b/site/_extensions/category-filter/category-filter.lua new file mode 100644 index 0000000000..62fbdf5ee5 --- /dev/null +++ b/site/_extensions/category-filter/category-filter.lua @@ -0,0 +1,59 @@ +-- Category Filter Lua Extension +-- Reads listing-filter from YAML frontmatter and injects config as a data attribute + +local function meta_to_native(val) + if val == nil then + return nil + end + + local mt = getmetatable(val) + + -- Check for pandoc List type (used for YAML arrays) + if mt and mt.__name == "List" then + local arr = {} + for i, item in ipairs(val) do + arr[i] = meta_to_native(item) + end + return arr + end + + -- Check for pandoc Inlines (YAML string values become Inlines) + if mt and (mt.__name == "Inlines" or mt.__name == "Blocks") then + return pandoc.utils.stringify(val) + end + + -- Plain Lua types + local t = type(val) + if t == "string" or t == "number" or t == "boolean" then + return val + elseif t == "table" then + -- Check if it's an array-like table + if #val > 0 and val[1] ~= nil then + local arr = {} + for i, item in ipairs(val) do + arr[i] = meta_to_native(item) + end + return arr + else + -- It's a map/object + local obj = {} + for k, v in pairs(val) do + obj[k] = meta_to_native(v) + end + return obj + end + end + + -- Fallback: try stringify + return pandoc.utils.stringify(val) +end + +function Meta(meta) + if meta['listing-filter'] then + local config = meta_to_native(meta['listing-filter']) + local json = quarto.json.encode(config) + local escaped = json:gsub("'", "'") + quarto.doc.include_text("after-body", + '') + end +end From 91b72089428aecca34d7964c10a2be7d5ada011a Mon Sep 17 00:00:00 2001 From: Beck <164545837+validbeck@users.noreply.github.com> Date: Tue, 31 Mar 2026 18:14:55 -0700 Subject: [PATCH 09/10] docs fixes: Remove old breaking changes/deprecrations from releases & ignore `.Rmd` in Quarto renders (#1253) * Moving old release notes proccess into archive * Copyright headers * Broken link * Quarto ignore * Moving .quartoignore to root * oops forgor to update .gitignore * .quartoignore * Update .github/workflows/validate-docs-site.yaml * Update .github/workflows/validate-docs-site.yaml * Update .quartoignore Co-authored-by: Nik Richers * Updated notebooks * Exclusions in _quarto.yml instead --------- Co-authored-by: Nik Richers --- .../breaking-changes/breaking-changes.qmd | 3 + .../generate_release_objects.cpython-310.pyc | Bin 0 -> 28462 bytes .../yearly_cleanup.cpython-310.pyc | Bin 0 -> 8050 bytes .../generate-release-notes.ipynb | 0 .../generate_release_objects.py | 0 .../release-scripts}/year-end-cleanup.ipynb | 0 .../release-scripts}/yearly_cleanup.py | 0 site/.quartoignore | 1 - site/_quarto.yml | 4 + site/notebooks.zip | Bin 62241340 -> 62247298 bytes .../1-set_up_validmind.ipynb | 2 +- .../1-set_up_validmind_for_validation.ipynb | 2 +- .../explore_tests/explore_test_suites.ipynb | 2 +- .../tests/explore_tests/explore_tests.ipynb | 2 +- .../run_tests/1_run_dataset_based_tests.ipynb | 2 +- .../run_tests/2_run_comparison_tests.ipynb | 2181 +++++++++-------- .../enable_pii_detection.ipynb | 2 +- .../quickstart_model_documentation.Rmd | 227 ++ .../quickstart_model_documentation.ipynb | 2 +- .../quickstart_model_validation.Rmd | 335 +++ .../quickstart_model_validation.ipynb | 2 +- .../1-set_up_validmind.ipynb | 2 +- .../1-set_up_validmind_for_validation.ipynb | 2 +- .../agents/document_agentic_ai.ipynb | 2 +- .../validate_application_scorecard.ipynb | 2 +- .../nlp_and_llm/rag_benchmark_demo.ipynb | 2 +- .../2025/2025-mar-07/release-notes.qmd | 2 +- 27 files changed, 1674 insertions(+), 1105 deletions(-) rename {site/releases => internal/releases-archive}/breaking-changes/breaking-changes.qmd (69%) create mode 100644 internal/releases-archive/release-scripts/__pycache__/generate_release_objects.cpython-310.pyc create mode 100644 internal/releases-archive/release-scripts/__pycache__/yearly_cleanup.cpython-310.pyc rename {release-scripts => internal/releases-archive/release-scripts}/generate-release-notes.ipynb (100%) rename {release-scripts => internal/releases-archive/release-scripts}/generate_release_objects.py (100%) rename {release-scripts => internal/releases-archive/release-scripts}/year-end-cleanup.ipynb (100%) rename {release-scripts => internal/releases-archive/release-scripts}/yearly_cleanup.py (100%) delete mode 100644 site/.quartoignore create mode 100644 site/notebooks/quickstart/quickstart_model_documentation.Rmd create mode 100644 site/notebooks/quickstart/quickstart_model_validation.Rmd diff --git a/site/releases/breaking-changes/breaking-changes.qmd b/internal/releases-archive/breaking-changes/breaking-changes.qmd similarity index 69% rename from site/releases/breaking-changes/breaking-changes.qmd rename to internal/releases-archive/breaking-changes/breaking-changes.qmd index 5436f8f06d..a71687d175 100644 --- a/site/releases/breaking-changes/breaking-changes.qmd +++ b/internal/releases-archive/breaking-changes/breaking-changes.qmd @@ -1,4 +1,7 @@ --- +# Copyright © 2023-2026 ValidMind Inc. All rights reserved. +# Refer to the LICENSE file in the root of this repository for details. +# SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial title: "Breaking changes and deprecations" date: last-modified categories: ["breaking changes", "deprecations"] diff --git a/internal/releases-archive/release-scripts/__pycache__/generate_release_objects.cpython-310.pyc b/internal/releases-archive/release-scripts/__pycache__/generate_release_objects.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..2627bac9f2daff392a7e081ce52e3cd393870a82 GIT binary patch literal 28462 zcmcJ23ve7qdfv|L?C$Jhv3OD7LliY4C6Ox;03}O09j%i@5ulDI3gm&16t2iyE-^!3 z!No2xGa!Yzg(4r%zH_-`pX2hY?i7Ppm9mv`?6OmFoXX2iIk8ht>~eCkoyp6w6UTCF zCw|1Sad%R_@9&w}T|B6}as|@%^z`&}_jLd7|Le-}^t4UC91mD)l?O>Hg!> z=kfD4wN%PZ*;*r2QCF>^$-Q3D<(jUffJ)6fU zZ>d-*N}18|uMv-eceMrEF#Ie5&)Glk+$b zKZ5In_M`T3TpzN>>~UNlwkPZpxE{41vmeLxnEfsG6S$V`C+(+jeZ>BV{WPwR+8?z~ z;(FZvR{I%T$Lx>U&*D06KW9IW>x6yEegW4L_Gx<(*T?cXVfP@gw$&Z|l(;A6IW&!aJ2G zUw@Kss@yH@mb6no`yQcUx~zr$cFT8~H>)>mjaA1(L7(khudOzGyFTahcHZ%;_2o5x zq1D7~?sRLl={s()_wpN$y>R*U*H4T+cjb}St+5HD&yT%%WsVJZ|D|#BJbvC8BuY~?b zh8e%^H=MA4uHn?0j$Ku^IlOSKWv_+VM(vu@@WO$4r|GygA2k~FrsIWuj$P-S@4VrM z1E^kG^;=bMb!n;Qu7#s0sBKc{kjS=R9dj!aj6VQ$W9geGU@i9?m_FX+J z&DECuRoAJuR{iBwKQsi)Q^DBO%*^?jlh!NE+O>wm?^~|9p2Im%Q_^yoSGZ;c16I9h zu}!az!LC`;JzzqLKmz3PcM!?9VIi*M*EKABYCk~chW;uRaYeUvEJ8Xo1eeWt=B5QP zFj#&3M3|9pkP?;bLwjNV83Owo%kHxuVIKw({jCG3;6Xrc11nlh7Hs^PMr*Ft@SYt{ zUJ>*wm%_|4rrB2*N~jJC-s-hww>9T@UYK@Qn_<@TZA?^{@qD0I*jHv(U3VT$(I?g2l7Q086ka z`$=pyU*AXzE+e;S+(~oM++%HRBV(uUXo9J_ZEWO--M}v$I14N#ehIGT?Hq8+#Iojt z-<(}_n_aj?ums!ywuI7J$1tm9pkB=v%yFF>o?i!`okr1%7iJTbNHPf`SxHfXNfBYU z#;nt=CI~Tl7K~=E+iS#~sC_1@|2}VokjPRsdwTW)v+3 zRq!3}*!ZL6_eeKGt`L=4hUL5vf~2x5$IjyFzJDNLeYa2j~wl=Td(VBP` zAxj#wY6R&cM~=9M@$RN}WE#kmZ7AR@cz(_Gy<2sE0btpxH_N*FFsg+HXL}1=xySH? zqCtc@AN z$VFaIJVIV|QHgB92=H{)@q84_&*RD}?nm)ruz$M6it)!>hmdUx0XnjoRYE{;;0#1M z&;$^z>keQvo*wU{1yQuaJKRS&oJWy_1xSV?#)u3+6tHT9sT>x)pa0#JGbc=)5Q^ z{Sw~u#*vHyZwinA%{1^vGj)?GetL8*X!=_vJS%92^gViTUl(3&Z*`8eF8(7vQScRx z^P9*JTHQsa>$(qWC-PrfFwKZCEu%1v5F3oQ(~H@~+#-05odMQmzogr_w$AIkPU(^= z?Aw`j!|nr4K-M%;i}|*(Sg?!xAptHHKLRZ!Dkp~I<-Q;49dH4XWofOhrPed(X`i3A z`~877WY`#APy+nHXiD#-Ihxl*%3RO9nc>^^0CS?^DZKhz2d}zds|Lt|PUpqEjQ~X` zA^I#SrdR{V5rN(kYD+)_S=BQ%#FDAj;c zmBLc(jjG?e;WWL<&~?{YS#_FoYgIsct>Ld#1~EhrB~Dxt94uIc7h}b%Yjh{U*b9Qy zn}-V>)T2sfQBt+Y-mLRz_|(~xmSFL8dC(;eL7rlN^)Ra*3r=+qG&G=G`C%W($#TOH zA`oWhz~jN`M)-PeZlP6&=ojXq&vIu_><;l+e(H@mXPGZS?1RvfRcvby$A#&|I-qn_ zQJ=6M())ZJU#?Wh>|CP`^bAeR!7@mX6KY$zaY$lI;Uhs!Xq>SUC%~|RQFtT+*$nVg zKLDxE1YYWC;O6^!M$)h91(aynKgka6(?+$DKCB(m%(wLSjsF%L0j_>nA9Iuo<3B3) z3C8{sZXgB13a~-Ug`~HhZf2u9$SBa=(i;YF*a(gT>oUYuS{i_%1~4_#hP^>B71jo7 zgTPc^vHN5@1B}iCL-UIT>Se&kzKf|XrA1W=EbU(dqh8MnmX_=Sur%9+r6o|0G_X|b z!qUF=?3-D(X-ke+S}xX(fonkL_98f%h{R2zqySoL)m6YIdZ4Ij3FlKr0&CS1z6hQL zaPqBVI{`xZ8)AukGm>|xP1e0xnk}r4S!bZMI0-ge_>+ve>bcHq6b(%0b;<8I+M#xt}vmo$gMKD%0zIjpE)9&TVq0|=w3ro&I!I%a?$L#b9`bm zahP0ZGS7q-6_tpZcn@b+(6CCvQB@S;2T;i)cQpznC{X5Ik;xbw)kgIWG6nm{_I7Hv zhvOM&o|3+Bynlp@Pw8G$RuX7cI=Bk(1Mq4nc!eef-X^#Ne7OxCcf$aNr0v}Pl+wD| zI_|;kaG$#a$so#`cc6vAj+g&-eSc~%_$gY-0vFjsp|=O%7_2lKd2mVX4$PLo<-!JJ zNA!xfC{2?4v9vQTI}d5G6nx)ln$aLG3UpqBo>qT@C;(}s?!!J6h)D{Pq75##)Vj%M z-m)`Shi(MIORd6j$E_($oTrF*tmGvio+x)o$<9EZJPvY~Umt(uxyx5X9SR0dT}K}l zl(!`q%jP5Yu&GVBPhtk>5f9WutWq2FT1RL_625vmzlGioWzvHk3RkKr5EO3~VdDw-s3X1hti& z^5^ls==p;=(ekw-wEcac{DWFHSJaD;(nhtQM5@kpQ7@IV)ACm?(A-CxO0^mms#S37 z)dus$YIS9`)=*E}{TwLGC+-xJX(kIyE7sEoNVZSnS zt_RWBHh+R>hmlbhXt5bXohHajnid-~EYKHIE}0ulGURV!qzz5KHZK#VY&hv<*Ms&< zlZ~p>nFpBy&fIU~=dp#@jtaZ!NN0>Y+fO-AQU)39=%#jbTJQ+-CP?u-Mz1DE$n*=ieK(r2Nj~bGvT!KYKN8#h zd#EK7Sb~~Pg&h-kuxkRhz0jkj_srxoXzH7p$vdgbq;_;qWI4O#Oz>kD#jJ~|;s7m~ zy<@%R_G`5{2)OnH?5|~+c}{F3rCFH3&TK2Yy-a$CAK!(%C3BLHwuJJ6+mE35A;|(HN99wX9|5x6JH|Zhkv=3$=zvO+R2~G_ zNx<>az6suzPdo_VZIK57GGucE5@cbqe{C}HAVB66&qbf=@lvWuM;$FGMc8DPd8Pii zp7Bc?1K^7p;3#Y|u;Pra=lnrD*W0<g9R>3pFhJ6Tv_af$%0~-TP5TaogP-2Jy%fuK}sZQGrEeHho~h_m{ns6BFHVJpSm! z@@k_oaqP1FXnE6sx_Tri0=dApW$gH7%zG+ZmP>3szHvkxeMZ z(2t;-nCsq${MK(CTy6nIps!akOI+dm5Ind zZRXx&@);(#nRJMttl#Do8epGN`COqrUQGflHpyx--MIsB5p9@azypslTw0D$EH z{nI21%P>=ckimT1CP~OBQqksoW`pk%Y(3@OP&}O$7%@tc-O!gsGel%~t-IDZ!U!OY zH=Kx0%!u)^OYegWdeWN3yU-XAm!&HWq-djLPp zTKJvD&!fic1N3ieBAfMN!QhWpcEQikX9F*d(w)46qR(-GMW6c&xD>fpY(s48eg?8e z77LuKy@=t+?4A?fb7wBZ-oRxR%?TdkuC-E{m7<3VFd~Kh)P=Lw4QDM{{!{LJ*V=ba zf^xShP>AO-w%;d1h=FK}?cfaH3!berG?0r|^Q3iFS^kqZVY#5)5|BzGI$U}?e}@7q z;{8j`T5nV5FHB9JI$Oo}R9~996t|~Lhdv1hTTJ?k;PXgQ(A7I!h!~9mLX4zR>TNn0 z@4b(^g50vL8-7VJc0sWim^a}%?;S_%C*hi)Q>;XA?<7J)pzqqTO@&K@&t>D3$4DT; zp>1;-rUAF18PAJiUo@SBIWXlaT86vqJ9wXkDeroH-u*7zYy*CL#p-fo>8RqxrwIIe zP}3~ZDOP|1Ux%!1d{D|3H1};h2=0N0B%+2webT#FhX zQ;jn!DrvZqzP38gz$q0SBW-O%-NfWV^A3ed8ti=_EzOZ{>)>Ghm!W9EI0BW5$VvlA zB&ZeajL_dtD%`AluvF0V<+?Rlx>^+dK&-^WaSJv?aSY8ZAaqBd;;IBIR2S-qD1r}k z?P>(2`z`{sFh9;bgC&y3CkslY(imU>lR%)43p=Dn73LjC0uq@-$Cr#E)as!DbK%m~ zMK9WQ_?T*Iuuj5<#095xZ5n&DJy>Y>2XWKGqI#$tIU@X%n6BGkN_k9|4L>0*Lf-y% z${Y1Dbux{R$a&3!B0?jDe;QKq!+4qzD^pfYV{-%+Cc{VBZPl9m2{F4%( zr`~|+sbda}=$CL;(iqxba5U zhQe+S52W7^fmhxFncp*!411}yYAvn8BdO^88i@=kj1Nf(O8d_33aA95Q4^D^aH12U zL`y#!6psr1tJz@ig&=3C3_{V;pa<@^V~K^HgocA@5ZP4(7s&gz4frw;E#y`5p5rvD z%rxj76sL;+Ms|1Jf(Lk`0wBqU_1#$gr=k1K(S=cfIK)bmpaa&Z?Ue?Aq66a zXKS$;p@qCSh$v1c0+B$Jg+Gz9!NAyBPc=$$k?cfCj*3BDRa1m?t7GX^#xO^soOqR_ zfBbn*J@NYFl_3An3)VyS zaqFQ=!N)r)NK9{K@W)<81jK?hSz9}9T|)3txt9c^gG9-Z`mej6MbAMg;uJCik@^vQ zDmH#fO*P=-7CgoXw1sN{qbY`X8e6&y4kuF8oUkl<}#xBqji1iiAp~f6w6m_UB3?c9cIBB)(~@OLYceHr5B1W zxA02iRwRZ(XC;&l6wMNqWv84)L^MI7Vm6oIt7mS+UvM3#8UqjNt$CGVtZ?xyH7NME z0Ya6MKmtl7ymG4(+xJJewX-F#N;ju^~otF!zThI=Hd;NW>DORK&WFGiLzzj`S}7n5@kbB` z0e!Sfo2H{zv3d($caQpX6xfQTB@2$in$+4x1pLH_aryta73Gdw04&Kzc(GMVNi^?d z-jeR<7;4Td=`eCi9r)B``_qrFUmlmsF}Xf={PFeja}Ul2&v!t7%ZQJuL7t39bz+C{ zQT7;mrO6m%J03y|@Ut<%$L*^2#;r)_fq>LIB@KAH#AkMw35lEpj3ZfEhVG~Y5U!~} zDwqOo(v^Xz&+2l`hve0)^v4Ct1kf?Wh+f|>rPZj;-y^6EqZu;?ph7rA0YJvEC4JNV zX%zR!8ua)ivc`|&Mlphz__wu19TEm$3d{hm78;XbuOU<+jE^Zo{zTG%dQakx5C??j z8GygZC=XVFya{2LL|=qqhU!UM-0P6Lk;e0af=j<|qiE+vD9$YQgJm!n9u8T*Aa56y zh>ZOc`gKM~_t(ziV})*>mWUPd2fI$1N__$=CfY|#6eVKBx(146)AU9vmDqmbIz(=1 zFtRHoE+l1~6pIo{ueCF!XrD>Z@q! z62zk%t58~Pe;^Cvmx@4-P@WbAda^PgdkY|Q%Z2cv5N^|a=+f9j*Tx>QgCUr=5W?)4v(O*C7;!Z3|&@Ze0OEksczq(5`ivJq%_oJ}&ONW}eFa79!4sJ3nd_rq}xmE#&1 zsz3o&7q->t+Yka+F1bI7!MlG4NocHr6S@R-L=K97T5oxw!I=%SYS;=Z$M}lyk2Ipd z+#lu}<`QB<0btzAGMX%9&tg~lWP?D&3VsG}RZ5+j1h24^fG81sA?%9`-<(7hv#|xi z?@`~C^mDz{fcf|riRcH=L8+JpCm~PSmxeD~xJfBZ`Op11Jf0)-i2o@0(e!^BKhHo5 zR4zp2kIk^MO9iE30*H*W85K1vr9^vGHyf4Gi~xVn_l%Jg!1gEp9?9jUWk`CW@$y-p z5RYP4Nxu-1?m8vbW(=YMzIQu~h+p%ul=tc_l=8nz_ytAom!apen?noQGmdh?s=_V`Pgh2+2 zVRvD+g}->12^2|Bve!TvKh=lf#c0H_5*AN|A`)yW`U;Vss&%i`J(Bp#F5z$N<(J1M zC*uzU79`Ue*?;sgD<~+PoU<9^EzSZbIM6KyJmZ!P!smq?HkQb~v0jaWwnXp(qj#5NTi5%Jt&_!CWcTp^h~tjwzyv0uVQ za|mh_Dqyfyn2}lSL^Vr97OMPH;8S?dj&ocY+&W+3z^W2DVxAoug8TDmWUF>Jyao3` za(|bIZ=vpJA7Zc|J{Exd{I94T{HmVY2V!uL0gfNI{}Pq9MTI2%o}N0$kT69F$JK79 z+&wT`LFJ=qpJ*PbXi&j=%HgZA4XTKDw7sawfGKd}OzlLBM6WD!DHJ!B=ugV}5FTUj zr}zVTlA69?6(?Sq8zN&MHibpte*rO4fIx_&9UE0VpIyxylMSK@keLPVV4nc0RT)bM zcso5<-7j?&<0V*sV)WiV=H1vd`&A=QeKO=rDM(!d?F=r$fgJQ&4AKIOcdDHX`n&H6 z5^h+|gjvYi5a2Lm!+vRedq%? zh5j{q*(Ns@cIkdMxa5vZ`Ob=Y zX6F3#?9}vRrGN7L=~rH!nw~v1d-gmc|6VsoDPOXKa&l% z@W631g&pkel=XNVf_pm;lH6FMehmiywcvqMo>!k28p~NzQ9_6eqDjw_jhlgfvap%K zghM6kFThwG1smbESg>ZJx)E`oMtlKjvT{(I1GHEd(Xi9Vgq_w`T8D}H{ zmJYGx3lLTvw^Ed#A4u(D?ad+O7GV_wM3Y>YRTZ})5bVOfm?gboBFuxRiV#GYc(Y$O zvqoA;VkFLkY3@ZdwFeP-V0_*EHB{_DBw7yB_)#kRQ<@0`VPs<{&60H0+(wFxp%Pk9 zdpahxNt6tuL_lt^!2qCp?7~br23!?8(}C&8Fu$u=w`-oPV-SU0?t3M4HX*aCJ5Z)% z3sk4Mq(j-bDDYzczwLM;eE$abN6_JcsNWHo>&Z9H8Toca0?XEc>kncDL{Yg6< z?CE3@a?vDwg_A(&jsM7)63g-|ZUmAaqMJejWe$=FENbo%3KYmGZ3419$TN%yLdeq_ z8N_iP;XXkIg4}^@zmoIy2m{0!SsRNn07fkCOBP`xrrRB$gft18mI@0Ms>2BNge^y^ zGghE`U1G_jSMct^i-=;Mi}*yRzhoV{P-IKnF-~wK;3qwMWaJ2jSeicDwRuLsBzi;O zd+Vm3yWxz$V6b;ebtu>i(Tfa@@noHCPiN?amyjcg@)8Q0#}X<+IK^BY3F5YiGvWSS zzd(V~PO(bEd?yh&Wd5fKRAR5{E5I0f0N7Q;hC1Nbru*yYXLsJ2(wV_BZLThTtF_` zU&JS}63Ju2W-K8pP)ux7N*M0mJ{2wShXZBD{ni0Xj6_6orCXhyebpm@4S=&n95(fWDq)Acw6J=LC zwml?;?A|L|3G5^`&CbCcG-k?%2m)dE0!L7_lae(ro1aA_+CfD95ZDsBi-C|dcR*}M zf+1hEJD@><61fp@qQ3vbDcxvH=~UC?u!OR}phd&|4XmE*TDdEz(TD1~NMcYsu>0V5 z2RJ!1M{Pp!7~Z7fjUqZ%gx-$W07Befy8i)P^+*#G2NT`uI&PrXNYsu{yhrG{Uc~Vd zR~`@;)ZODzbb!8y&4ub5OMfNu6?4_?RBQ_(aDPffb?!Um`_RGIzZwv! zEy*{j%w2*5_F$?dI6y*rc`SxEqH>_R z?m8fXk^}YZXXUUJ_boQ`B9p(#iS|9 zC!grp@tXM(K(A_|HG&`|K7wE#S5rxRD_cSEcfNTLj1HL+L5a&Ob|sOpf%I3Ik>jrr zc_wXtGki3bt)Vr_er*Bj4^cnE?RY}T+L^B|HkYIT8=wAatFk1-MG;?6afu&;!@|H#}HeW+9u zB|bjR>-z~>hfvQfD9|#nUrDLFxNaUWg|hvb`&+2eBg2Ung+fv|@8^*CRByHKWU!sG z6n(33Q${5|PM5+W+@kPYYurK&pwV%q-6v8PY_SZ9?FaruJmbM3PkkP?A=r{k+>1Za z#@1>$610^g?Tq&vkLQV>tn~Ya&_DC@08qSx&7rhEeJ*ufj_CTV*_LB`a6n1+^C`6b zDOInHdswY-tWP;#`>|LZQHBmrC{1)+))uWKmiVaZbxVL;vdAWyuyRmNgtA9OL;>z! zg;noPM}aXNHZA>if;GFiKS(-E+(3OtDuwT^+cA8lcl+@osi8B9;GfT*`XX+>Bfa;X#O zBGxc~MRJy|;oK0xJf#5zqe433)e)2;t2i?oW~7H6Qb`(5JNgKIf&_JFB3JrIA_Yd^ zGY16X=&0}OM3i+{$Ge*Q%Xm>0hzq~)3_qXK_Ux|_uh{+}9NUHI<4B!YW$NN2$qVH^ zm#D<^jNmBVReuvS{{*`wn$y=wgC4ej{3>q;wk&$3wAxf_2eTO1YDdNi&A7Z$ir@k9 zwLCOX9m*n9A68%(7qH`1znx})Eog=^HOMX#aTOY|a}jPs zXnk-n9`1H9s#1id8Ea?YOoXR27(zR;bpdRQv6T|^D=tN2-FOoR9^p6z$;ojFxf+jC zP`pNtw&}$3DqRoa2!;C2mIk_jcH~Yw`1~$(1p>tWK6Qdl2aUHo0}_YFOyrkPBZ)l( z#A+zF3f(&GD9C`dG^)3jH1jgLBoe`rpk5+9OPt+S+-~B_i6ix7R>Te!Tc$&;CPrcb z_Kx2~jZb2Bg;a`s6pVBZKoM`8Jfi%B@S0!1B`ipBReWl~`$Tf9sT%au3HpUY+rG^G zB|hBA4F{94xN@TN0e^!)OrwQfj708AzQ&w0aqR1 zx|bsS21tjv;UGaNDJk|@+ut0h0^oZ<74lXOsHWe9gGN9VIR&c0Kj?u|g5~P+ybIen zG$ys4jK$y%r3Q}G#MY)cXrl^#K|{Fye5Z*+s(R_?PO{cF3C%LH-2o>!Q7C-+)6Za> z;$^ow0V~yo)~zud&J}$LV^yqY*VhX;wg<%Gp|jBoJ5@^nZaOG;U@h)qr#CPeBEd%x z=>Bu`p~S8Z=o#W9t_(%(R8|3N?p$qE%5D$99 zEDDT?CH5+45ugr)5TaFT1Hu!+x!{V!$xzW&JCZaT5giEP0yIiI$6@U1&ETxH^c{*l zPcV2;YUL${qo8E>JmW69Pi_Ybq5_1w0% zuk^glu@&*gW9wPFe;w!WuIGK61$UdOumv|^VZ9G6ne9H}QoE0#X7&Kip1W-iVn1?W zJ&zOOzR)hT^C3 ziBQ3_svfo%UaPs|Yqh0@3NQe?#NZTT4&#y!iZSMX770&5 z#W67sPWC9qECTzc zlJW6)VG;H~9B_oJSin5Tr(#JmxXZV48MV0sLqr}r#gGV<6(hrs4yx(FkK zt_^-&*A9MA%oZWx;hHue;|(GYX%7Ff6q4Rs`d=LUT47*RcmEPi^pLL-3KH`5DrT7E zOGp>AGa+3(C`j#9qe%fFYS{Zi@{k32xF691MmrOct_dJ)*Tw{H)dH&L6QTvXSNFXT zt&eu2zFk1Hfg)Vk#2|zVn^uKzA+A*pSs>v`T9g(l-7Uz-`h;+0>>>zPKL{5NyzC)d znT>oxxC$U#B@n5%aFAY1xZaKk*W2h58!Ciwp{J55XH7&+p{+v~im z;`32R4E-LIn8BK;5WUnn%Yy3KF7&+{^E`4N839 zh4A-&p(650#BcIYk?dhh{>Hi_E1}r#ACXf0FDCyRNnrI7<#Fp`FF78c4n|`>Azdim z0I95-P9NTpPMPPQFsFTr0wAtmMEMiv;c5eIpxyer3vgJS`j5oEhpgbd_ z7?EJxS5Ykm%o^_6$>6 z&si?yzoaIPV97OBw0;O;cS6)0Vc3VHxN9dm<6I`^uzj~KB^DH zsHneHNb7GM`r|!A?$^;tj~Re^Rl+8I6E|ujAEZGC#0dM(+J;*f?TT9%7!0Prqt9zN zVK^O~Jp2+2dj^bonf)Y+S>$lC9CENtg9PR5LUfl#>@xoShvcpwcO|()nB=?xBAv4b z_NUyJ+U5q%zhPeWCdA9ci0*;(nI}uK2SM0hZ)e@Vw(_z)w4QB3=cBBR_rzGa^76_) zusSd>zPn=B!yr^6!Oy;?Vv{3RZfs(Uw6B;PD9DuK#8u$3;CHXRTL{}2%wirkOF(8M z)nEv*R3*O01Sm1&bxzUQrZg)hFA*=5w8=);tRgNZHoKn(7P*05D4Yoj&pg0@Tsq~H zaa)`p{~4>Nj#c~>=D4zfc~1T{3dDWUd==#J+Oq{5^ld34`F~->!$@%6cE;_5e|>`G zB2#@0Iat6&uNRx!|Ky{iNGkDO21r8DR3=Bx3+h!y8rWVYGF0_e&8uE()S5Tk`}nGf zlUAh8PNI_g`>67_h!W$-3>FMVP$C$ApI$-`SV=R!X6E+bo{|tk#bDinwd*en1^q4K zcjetVyq5g&yEzr7@$>!&tSb15+W)nL(^DvkaC15$b|c?l8#}_@*2w8;)1HS#>dTOu z)VA}u6cL$eKa2Mas7_+VQnh^(W#s>P+29EtEkK#KAv@q;CutFfbDOs5_uVFN=if=M zn?7WV+ZvE9zYev^o5epdQMj$~pbw6~FSUV0(k4{0ws{+twCKq|JM9nNrZlB$4)u&> zxaV%9`wsF`pT7t49psPrZOBta{~nY;3*H%KAHH{@(7E{Cz!l`%xDL8O)9v z(E*G-?cXal#Dnep?POm4`=A)(-vx?$@(=aiLlVI;R`@lBNzdA`B4m@qOx_SCD^s@_ zqJNf$G{XIEK%$Zj3fzM_CYw+N3%6~iv#xGxR|6c`3QC}kgpdt7anG@m8py_@C2lA&BO}8U6kxR@&zfI_{w?U* z2_a|eFz;K(a6WO9hDxP5r!nTDUB>V$2%&o#D>Pj`5@yx1x$kquKF8#3CSnr(LFT^1h!q3&Aw*Qc2FEROHB#@%c>>5khxBSbb)^DJS4$nhY_Du3PX;Y{(@A*#v;fxr$FX_}~S+k^#5ARjMR{6nKRLlqdX5Q&OL&1-mzLMuzs-UlWx^wJT#awhKB24&I4eL7 z7J)g}Q(mLo*$eWoX{jv*^s2bB6hFlr8JIfS?Iv>_5RzpS5D++$#n}Q(vJ?p%bwSQI z4>Vigh5S_}Or2atcu_DU>nqSLfV7tnI>|oVhUPN}6UM2p&thIu-sh3Xp=IK@rt_M6 zv1!qlJ=r+kIQ7>{pEOUK!}@)U>sJ3|H`R&a@tv7BOaRLPbI`aK*J1O1i8?b)KG?ot>Ti!2$&T6iOM1w!pQy0OgePgU~b~3Xmv~unT~!a7ET?usr~m zTz)A?e-&o8 zre~)6_3PLDUVrbso?f}^X!u>9`JeSWf1_#tO)sNA11~pm1;0eWHO@NPhSty+J?kCT zVDhdvba^)#2JS}3+{iU@DUH>ze0?X+&0W^8d5&8@)fxq!=Qi%ny5<#mfjf9E@ggtb zUgl+9!M(!Acop|CKF%j_uktf|68CX`f}ee?i#z`KHp!tc%d2+oYe=ZoSYmT8^uzXs zzlfLOx*s;%{jKm}uN$4e|M44d-v98!YxN&K_|AuJeFo*t`X>)s)HD0(crbC@!WGP- z2(^&y=)2k@gR=|THXCZ8$@N(KN6hrOvCaAuu{LC3j+-$2JWWU4T*os+`ZEAy{EqGZ}h`0DFpY|yg?^)Ka9)=wOKcM zX{k5p@^*LK9rRt{gFNs9qMUUT=G3gjfEWl3csnAB0J+FWTL(rbCg6 zZIfDs4eBp7QN(SpG4np=AIz-!t;em0UU%lLUTd&{*#t9hx5IY^t20l%PMdGEyL^VH zOOo7i`yeOQovmgIB6bJ;>Hbz?lcK4b>^Z2Sn6P!jVdfuoUH_(PnoN{Y z*`n3Teuq?j6SbU4wd&6eC{%j{wa09iiF2XB4XIr7K)L&MslvGhh3CCV3?TJ#>mgs;C5Tds0PEXGS?arm;tuoWS@naL-V>8hh34I{wRAf)3BScrGjxch)E_b`W z&#@+I*9ae5oiEnRMDGQONzx~lT4HGjHC@_YQcUfI2&5^hq6C#Elfn;0+kfgy>uIj_ zI@}kR&=OE{saVph%w$y^1~iF^Lk2`RVc{}x1^4RI z7$I?kX+XbRm|NC9XKRd`Joi&O)OIYQeT#5L|6D`8MfI^RYm6`-16`<>co*FAp&jP7 ziKG3D@Tl-S$DJKL%KG(CFTFG-VGhKk{0yh|}W=d#7cPE->_|Oa8 z4KHjdBZbWkT7W6v{aaey;f%V~U&u_Q+4hCM*8Z4u8Gr%U5-?)u5n;BZK?kSVZ3mla z)}3?dPF63aBO^S{;;I|;1dPS?I{k;vXo1af%9bx5uXTsz+_^kh%#k2+z*o z3a+78*JAC#WTtY77{)U||E>urQbDic%Xnx}XNkHOU)1s*i`2dI+Xis17^f15R1|G6L$EHG-E2nY= z4{|EiY$^nMJD3V=3U))2iJ6*0jqS1ep()s_a1(?(i4%mL2m`_~aS=RgOm4q}DTtR* z`w=cF`xPy+7Rckk^JMM@HbDDNdL4y;kumEyB8N`VYm2=T=!gMXjDlvEoB>t(eb;O8|k=F%GQ#!kmQv`1&wNEW6iWi=~MxY#0?=~ z<;)3MxedMnL5o?eFoi27yI@Qhwr=X*Vy0R7#?-~@AZU?pvmec!CU6T^u!jODBjXS0 zlE_&Z(Y@4NL07;Ta#uQZm%6JD5S0K2ltVM1&8wiWVgnTU)S($4XoM6yxmb^lT^-0_ za9dIo6h%}Htw(w8#KyW#k=j3i<^wI3{Y)*iw^o@pG?^BSp=UWZGrajgyNA(-<|n42 zRnMDj6G0Yw#c8h!dR0#9MUo%%sf~HR$dKz+uTNM-VO5BjrUO^tt|5qWKZ2)0DB_E* z*O|$hvY4Tssdf6v2LCAab<$x6eZSRSYXc&naOJMA!raKRfr#N%>3-lWC^D0yN@tn+ z+OEF|U)fR)Ns4hQ4q}_io+DmczK^)wDYAds2Ly*S!A$4_(G+k|04VK#gqeH?vv#-D z8F2E=l7(EFkvQV?7;{vb!}B86m}V+XI!%%xs=e@<@0Aa!01?0;&V_wGc@qh8T2{!u-cAzfWOrEhh+{6&{r{zS4S}cS+R@XX~1}rNGV9l>6{jAuJ(Ez zWLaVWOjQ|{9DNfxS~@o3&K7OtwE;rQtL_3KFl1v`8nM|9kid~tf}$-6atjF)i&5pW zyK;Nw&V2pu+|qmVOM-HWhDodSH3b1t^|E{C_VUW@#kW(6=vebj?wj*(rB!ad zej{OG5i-OnfGLPS-^nditjxl$Tt5=|Vymzn`O8dCtS@<{4Vl$0h35_$Oqbe@hym)Bjhh`4`;#Vpok z7zME7hJ-yMHg*iao{0@<9)i7BMmI9g$&G2rGXFd`%yD~Yg$TfT0epToL|`_|0~qbt z1SmF!g}AU|0uTwx-o+TXxIl2hi^Ker*|e>Mwx_Wz!q^HzqCtq3Bd~ama1dZfHDmJ0 z2G8;kaD=5Szx+ff5EgC=0i=Q8obo@ESeX5BsVtdC@5@7UwX^uQw!B>#xs~47J z;+P~~9$UqK8(qaPzk{TiW)v3q8M&5|Dv;$S&S}m@U#YF4BDv<10TP$q^wvgaAGosc ziGU4JkVKQw*hj`^DOCK}6W+$rQ<=Jq2C`_304l)V%MSlke5W8m;=I(wH-ZI|0Z5{W;9-7ZGh$?A-6MvS3COrcm1-$xm}dfB}@ z_uhPcY5vaq-12;V@kcB3%k@#rS=qaI*}e7t($f6mO8wsa+){mhF~g&%>ZZTCT+dlo zC2*5jwTo~`Zg)&KlJin6;uRXabU;>jBWsRxEL=35#hv@XF>s%>lg+|Cv=B8%d>70T zn}Q&M81d(L6@Nj+Ur|F&f|B@as-T2Cv4p?Td+Mu`qoM+*GOelfKr+LfpkS(bs80Mq zGE;DsC5^I2qfE+gw!4pzP58VqhEs=4-|sdzJnDwElIiQ)W{>3Fx<1dz{W7x0e1ehzKa!`8EU{j*E| z<-9(ri}%3r5YExr$Ue?BB+lXJA4lFCXxBk1%8`I_bBN<_oR=E0snW+72U-&Ma&eAO zWN2Z>>W6Ue_0b)Rgp`mbFH0M^m*<6{&2{{pU+Tj`2>j$la9jW)mWB?lVwj5&;sCMg z2n|caa$MfAfP`hB-F;pGik0}-c`YvR>UHhO?YI~`iSvpENc9AA?ZLvPERW2 zqK_A`6Z{K5i-7AKaOr&XAEUT*vSsf>6q<3$gS3}+!6{qu#OO*zMs!?1@>K!5>F_xm zaV4{M5pYp9ow7O(RG%C<$<7Q^VvXG00%oHv3UbAQ{G|k-XnY?Xj^j}DS_T5AqKs41 z6MSsw_}E-BTu938Ea$bqca~Bmare}TP>Oz< zszq-@DIqFQmMQfSxpyUkNXbdgHF=6QlITY-rxZu#qU?p&iV_=DkT*|@a)6D0Nv{Nu zVgW_s{KV@Fe3=W2x2cx$zFI}1;|P;?sEN+EBpix&@zkiO^`(53%%^dtH4r#YQJ-m2 z5CoC(drE`Bey3C+jj@zTrUmgs8ssJwM^Lg#8dQh)s?Wa@K9XQ|!3O4yV5DI(2eBYw zA)-BM0ZLZsYZ0D+iiVDNd^!5;((jy?Y~4*!Q~U_bTDbOq6s*(noQwhg9uF!JmZcE_ z2X^Ump@x`DrgovVO`i($^x+aoE0YKE>a9R;h#l3NgJU?htq*bFPqeWiz8@BWt*Ap! zMN>M|zdZWJahO+Qi%=6eHg(dR$ElX|VR>&gpcE>xo6=1+n{?!XR{4QU>Q|mY%Fi7# z_ZM@-g&c!i0eo(8)Q32EKux|uo>Md`gewQ9&_@+yZg}|EfTSjP4P7STkSyJTn^Kp9 zyGeD-E>Vrc#QPv4DXCwxSvw;?Jt|+V3tzeS{OuphkGsMkG3wL?|E42yjackgq~z2}xESqbm=$LK?g#_4Z9P`^S*g7?plx*sz6 zCew8-rCui-QgrxeGHNvEuU+|v*WviXzd{!9{zIiO^Kt#*aK5ycoEx9Se{laJKBbq> z_>X{^SMkQotN4ugP~y^7eVE`9|*$Klqam!*iF|YV?v$xGETF3Tl-LzMpS4x`*f z>2YAdff3^_%EaezIo?HhS2@*fJUpL_|GuT18860VXlnK=R8z@ z#zeh(&VzaN{7xSJWg4B-JFl>evfrCzhjul8pU)Ws@44g#2qRlz9Y zOF8DJpu%5y^rs+Z^rz5c@TH_qB>XoLlS?EF`gg$_c!ZVzMiAT)MiJZ*!4TXLm0-V& zlF<{@_$xox6Gc7O6T>{$6E7CKl*Qx3#sA3vb!XD!#Q#3>Fb8qs4F8@caILt?{}OO{ zEGGmPue-6};rZp@;pt@P_}ydAs2vt3FOUj)eTDfy?*i_`ME!f%1BS&N|0Kg;hsEhY zmy;r17ynx+>fyRL=Ha@;O5&v??|1L-@Tlj0_b|`>B;C3%r7H?jcm5>%bMjK$_-I4Izthyg1q-T8Nul_HdE|AxMbP(r+Y%M4{Rjsu9IP?1+BVw%z8K8v+5A09 zC`o-A9BeU?`gU5ums#S6?DYR+$-upOfly~dc9^pv`_Q+S#pD{xZRuL#KUI=Wb1Cf3?;wE8Q(*>FU?*>J+V*+5K$U&_fI&VOahk%#j- z$0fa->3r``)ZZ&X^v^S<@42l;hH93x>K_@CmE|0I1DSKO+WVz1o~=1JYyC&cXnXbc zzCp=*cv*1C;pf^rolQ3?KK(WVF(%kZCLuIgNp_e2TNgRHED;`+}8$!3YN~u z?tid+=M>c|j^48yu89<%dy?9+b@oO|`{LDt%s1Yfd1l7n4uWF#JRO|e_v(_mj<ioC)0paE*Hg8XW6@hZefTB%xj#U$ ze9s;`3+5z`J`rimB4{KF3~FBUfUm8P)LKaUQJrleYcK9lM;sI$dzzD3E|<@gap{(v zecv7vYRfxRa+@gdpQ)`pR4j|QwJ@u3>~#_m-`Cgpt`a(Xd^*F-V{kMSUxfl?_qn`n z@~N=lnm?y?W82m3T)VzJZ^|t2;=LbWLVc|v_)c#k^|qM~VqiP$EpI&O4HNn48H-v| zUh}kb&L62-!5Rak8fW&eY4A*p+QUh1gYX3WeC!ZCZucHEkAz*tKm4gxPg$ZVR*P z*~kgA8`xM2vm4oj3bUKo6biGO+O!JaU^ll}5N5ZuArN8D;Hz+kgFUvRQ?zq~7WS!E z3&il-VVqj`J;~HEC@Y-pG9FYov;WhOW`tHa3uG{%ok<~!0=zYQ3v#?Q`wP0fH3tjU zyfwcTym&K`E1X&W@wk(*{g3Z1+F2;Q(wUJqte|{B8GV=q4=#@>KrehjAHIVJzla$iAyDaTo)K8-OiFvW%y#;VDmaL^r6t_*QL8)Oyiw+ldm4vq>=|yA z&dmQfQU7fyGJq;)g^cf2&R73$8jDXm#EX!KpE<3E23Dsve8DErmOF8lkdB=jDKCkM zbCUk=nv7S~&d`jP)y}a0uZ9S@pzZZTRQ#}r?$m(noaZ_+BTT0EljX{H?Z+w;so|Qc z|G%5K@IIA^|KkSve;)GxF5<^{E%I>B==6;H91o1dkCh2&pZmCw_~@t6OI8=9R{Q@= zv%>14)%yRiLH<9?^?#P??I7W{Pzcgg^ScH0+eyN0k&tQ4@4?h>=LvVjKACF$$fth4 zPPilSX935}ZK0|*OKo_sk*vy1j$C*Be^CXYq zca!7qFz4)ed$S$-8-Ok<;lya1s1xVOixZV_kvNeaCl)~$BWY}qGs!qrH_lTVr>fyX zxjNq7!Ko$C#SUC(7EaWQ^K{3FTDVZ0D1{TtaD#$uW&oV1ALp4Eb5XMGmNWfkTPUt& zPCKOQ7Dn(@&kg@=d#T0GJ{xXB2OU|>lbeo@h|}NpT#q1}Ezb-5ErnN5vcXIF2wC0x zL?B*MM=Vf6lT7|`+UUOktu%b48Y(5EC zijBt~VZ2ZCYOTV0P_Lt~ECH#g%^aJr29_$}@o(IDa<}^f=k+49I4^(mF&{S$&Hx=>Q%Q2~fy4xz^tjo_+^>p1-#o3p$ zQy=SYtBP|j7pA7@x~Yl7m(i*9y4z~vJj->d!@6$j;&+!DQ}=YY)y0u~%dM%@dTtux z0?Xa0f_mE;;zG+qsd{>Dn&KkMQ>l;jwl&4YmKRb}^xU+>C6=+N^?KV{;*!g|sl$41 z+Tzm7$EkaI+uGu?%LHlE`fd-z<(5g(1og2G016OX2cQ8VbO0y_tpm`3usQ$}2(AmT zf)Kg@CkU+z@PLq5U4Rb+*8_w=2t7aygw_KjL0CON76jJ^6hH`lKm~->2Q)xfec%BI zZUE?m5C(uT2yFnEfv^U^BM{sWumvFu0Y?zp5O4uu4FL}j+z9XnA&h{hAhZz>2*Mfx z&p>cvARL4+2BJV{V;~NMH3pJEa1$hu3PP9wnINS}dT2Zlca2*HR) z01+7d2p|Px9|06#xD`MHMpyw*Fxm>B17ocKCNSI@UOgSuordPyi!r0TnRX7SI4=ZGi`1xE-JmM%V$yV6+`z2F4=ofJb1s zJzxt)*aMDWv_0Sg#@YiOV7LR|4MsQsPr+yhAP|go0G@&2jzBmV;Rr;5(T+eI80!cm zf#FU-Dj4AeWP;I7KrR^T1iS#l5kN5*fdI#`~Fv11s1fyMmUNF`L_zH%*0wZ9AD=+~@y8<&{tSj&x40i*Tzz8>B4UBdJ zHo;go;0GA)4(x*w?!X}!?GBuRvF-pK1nvP4LJ%GR5d`f4kV3E?00ji@3D7_go&Xer z_5|o4SWkcn0`~$~AqX#k6M{s00Xz__7r+OBKL&&#h{u2!1pOG0gkT>7vJkj8pa4O5 z11b=-H=qH*dIJw2a34S)g75*1A!r}K41)Cm9zozw09y#+3E&7pKLK1I*e8Go1nvuX zLlC~eQwZ7@2!vpLfoBl-Qy?6IcnU;8&`*Il2=*xwNP@uqfK&*=56Fa|{eWBu)(?0A zf%^l+5QINa4ng|^RS>K{@Dc(K02&~O0N^bI9RR$CU;}`U5O^T)8G;A|Iw9ylpcjG- z1inJxLBI$E5d=&?&_Tcq1RDf=hrok@B?uxISc9O0flUZD82AB!BY}M+1c3w&A!sCU z3c(@)JShAbKnO)V1BjsLX85Qazsh+ya>fE0#J z0w`ecWPk>SNCu!VbTU8(!zKevF!*zT6^2AS2RLEq=Kv23`yAkd!BYSs7$OA_gP~IZ zNf0PmoRt^&;Uc^0B>RF9N;|+n*)4=!E=Gn zFhnlU2}9=sy)bMp@D&En14dwoJYWKb&I4v(*gW7n44w}x!4Ubt8VsEeY{Iblzz-x0 zUI6UF5Cy;?3|#=6!mtGZ9xeO@KuC*t0T9umUjU@E*cSi=EgS%7Xb}Jar9}e(9W53B zm}udJ04pt`5a6Um7Xm!A*g}Ah7G4Ah(IScfFeKU!yyv zqgyWm-v!wi1VU3yE_UaFPp$dCCtALLFW7A~>SJQ|?ZfNlAg_y#m2J~ol8mVZ-Jc|* zgHLZaL1~7aj-8G-{65T{ufcQXSWIUYm#4=q&uSXo-nlZ&K0yX0W+1sD?NHULiEba7 zzx0eR?^_SGR2(nrpzf0~&==z<8PM>i<^@7zeOG~uk$%xbyo3C}p{R{QIgw0fZgKhd zvUdG>@%!Fy32yeI4Q;Yt63O&&bd$9%X_L#v(%{zk@25xdo}+dfWr=QQ-pvQ=7uRyM zG2e1R4y5dmZ-NbipqAg>WVU5Yliw@(`T-VbqOWLp@krQHri7=F?w4`~FSa006nO{R z^E3mtIopG=djCC>vaQv@UpXdc2z^qw3k$7! zvjHqvp=7#U)Xcetx97l@RSINoEJ{Rvn0Qz7jQYK(EmwZrfagX`{) z_V4fE$TEYV>9TY0^wN&GoJ3jFsvoUX)$mbUcB)Z|mqgH`#dNyr)@Pig6KtdwElzv6 z(NxcT2)ri>`~1cp2TR=9K-STene4GAD7P0|&e|^~nt1XO4_$vkdF|lLvacp`xm@k! z&G=dD#U1y8maLoCmRXlSZ3unYNcB zy4a^^cCkc_AIEKc^a;E@pS-^uH6Q(4|54@IxYsZ&l~@|k+~ci%^4@3;<2U6ew77Q= zr+uqW!e!Jxs5$?@sj{um@Aox~(-kMG-!i`-Tzpy^$Me>V&4GO~X->Mxi)k0hcN%;5 zg23BI+crt+=HL^}dXlGU=XnX_=>7UaSAr!Sf~8q%rbH^Hom@P>RJ@~R;v{9_q+06l zlP?dC-e=~Dx__ykbGUeFIQuPioqXW9$R2Ky&KNuSLJdQAIt0JF%XxPf>1Nh|ROMmV zEsIBmaWarSi9_vJy!M;At|Lk3@P6{n9}@hXSP|N4K3xu_HS6z}4be8z(mzrrx@C%M?NPO%oX}^r>hu z^f*ZK`kBcfowp048x17Iv(YtEI^`@Ul{P12khOC|=zESTKhJaxCt1;ZvCVu4O2>lk zNq=k`CokFWzDu*S)J|sQxvzfg?31@Lc<|KYozGlin5XCJw-1mHMNihfpd~%?wV8KO zI*oMFu8nkx>s}4=je)aB_g|d`N9Tp*LqqSSL&RM7;B<C}NmbLFU?Jto z7e_YYYCQ=Oj?z^fIsIQu^No5}R`ao8IfMJ`cP#6cD>~PGF4)hkP0p%$lqWR@3p6(L z=!+)#g7SyHury_*W}L-(uhl>Nq+xokUx!9aId4FhLDBE=$QjQv@}K_6zcjtJHJxIG zM?)4#J8F`Np5_j<)j@iV+vH;Sk`4{N@-51*r#$p3mXcy0zwvv}wSWA1m!I$a$h;-j ztl_wvPFe5bsno$283%e%m-@4(iw|u^>4Ir|?{XC@-v7Y(wAnB3_Eu&`edWf=12(A_ zt9#cQ9{TcFAR`5we^F0rimNt7Tk$tKekc`7C_VK2E!^%Y^d<3LEmuMjSMd8C8kv9C zFI>TNnm-rMcoke~ZcfRl=Nx>|Zd#OywmcF~9FI?WeOfiSP~x^2|?+=l2d*bWi2az%S>XN99<( zR4(qAE(4QlZM`a9=i;*6z$5)ev(IADt@FhJVz~L zc|Fye-43Ahq-yW#y22e9uiZ1|srZ`(mBJ~jl^;CmK^xJzewSJIqJMfGW&L31?(}=Y zGNjIv+?x8LbJek+4KTE<^5dbrKW+Vr2atd0WWme(M5yJtd8x1O&;?%(FFZZbD!AV& zs7p}s6KdMK=jGeC+lk#S55Orp^w2)H5?8rgC1TI1BcgdBlBnT+d{5e|pVyoJH7l^_ z^R?4bi)!@*w* zYfh=mW!eSI4my#!TS?5dtBh$|yWjhH+e}hf94hHdNRze%L+!dI;p*QsI|`07d~kPM zG-bDuPIz1?wozPoYk=DRv8F)YO_bu}K!J0y7M|yCgl;9Q{9Lfwb@7QPc12d@D!%d> z3!3)?FL#@)nu*7HExXKJnyN*eE)yCG2Xex2`jQdQ%8})i40Fzj!_OJ*B3HWSJYSlT zWh*P~*Z1f3Zl8>ox9YMrHNYWz$7h-(1BS1kztmUVv2gUBX&kHc|FmG^M}Y9$uY0}! zHY?7(BV~og&XDxW!K+SW@!I&SA9jyvhE9RWPHcq9ibBfWcFvRS={G)=FCVDZ2G~+O zfv@|J-NfY|Lc93oYEM)aJHLAV(nNK~sV?lLg6htNk^kpAz2=C_%F&NyyJueGGuY>Z zo`R3f`(~hQ(+<$S6d%jI+0)$R25B*_%0d~H>+Bh~^{)H&@VB^)oPd4NMS|7yoBeNFC*QTfGQ^y+*k;M)qtvrE zX*mk+Ft;wSHwlTZ-ih#2oll%=c0WLe(`${f^&kDJSG-X;OhkL*Rz&}@>R`22`vO&T zU|Nwykc+S2P|-tkB!_C3u2$qhvg+^=!rz=rerW!QKlduH$+QcRbKd6Q1}o2#p%0U) z##^>KZX8zcQrVAq8ZFV?Fv(Z3K&u&6MsHI&&c@?h=jJTJ zVjIg`mtf?vB(0(wx_Z;RKGaVHOfm!@VRn0ib@B5Lk;v%Wmw8={`A6ehDY-d$)pptY z&Lmb%Njh^ren&Pseb52)v0EQkbi>b|OGk>0ZFWzj^7k*K84TK)P9A7XMXF8o7Y9hC zye#NqxT@J2$sS-NM0_vXB~_>(Xwx#!GUKdr13OsKX>q^}kXnf}t`f47;Kt!P$h#Ksd)=N2$FJ@AWGR(@$BgS}|- zN82flf`rnUli^r~to`0jgiHSCT+4@fcLyGs=}-Ec@$rt4*Q;jdcp4z*gwNz2E@?>r zs`OVL@pq*Xk)13eP@BLn2_Q~=dA0kS@Pk$c&Ra;uuZ=Mu0wRixknKzJ3#b~z#4)1# zXrXz-GTSg^%hH5wx5~~f!oVQGlumz^JICkgyl_(j@?eu>1HCz9R9(MpZlY1TIN0&9 zWHgrgmx1_X!pnBCNsB8^sns_s-N!9!Yxf*InjTV>yzaHW*&xs8h~|^OF*h;^wDf&-V~c#H=hXpd<=$%Kw*<78t}k|@$$oCUEUIp<*kO0t#n^Cs zv1e=$Wazm54PfIL^lJQKWj0Y{w{!cBRZ4mve5?Oed+f@M<^!v_x#Nt{gEl9hsbgdk zFVD>GgT-IpnCJ3pfB8J^5(;GC9Ff*J3_pALglobrTzkX)8^teY@x8TmvGA4tMynHl zzb!wVQgflWF@A3j2SbizhW_jXE@~0PWRb4=L`6wJMfyvFo^P)owp!eAFxM+JcMF#? zV}LJggZ_C#k=KMNi08Vo(QtIvX9|NH{uwy>%RZOE>2>Z`SBpGJm$8)%`z=JSi}ZuB zAlBtxe^WpQsr*#%HVOWX9jQX=2aKUc<84g`GyB8)CV4f@I`4m6UGZ>o9_FcH)@=*@6(WC5F1la|0=Zy5>u~G zaDvB8-qUgb@*eDc`Sm{7C9fuH)rwej zt|&YQ#WwRkfLN4V{q4~=wi!ttn~yhy=DAImCILN^K3K67fsyR!1C zyiuM1_uSkE_tCUI8y%n7j;%L3dB1f(3M(Jw>+OETT!_@Y2!3$@s)bjYX3Z`zt?2*$ zsc-YZhg=8uMci*u0~`I<&O0KBB_rnLBWWptEl-<_z@PQ8_T<0&B~Jo}Wsp3JD^01M zEoXZ!-UkPVtA@3*#_hPuX58!jdrZnSohj&M$V2a0=+qCiV8{+Vn)IWvs+P8M=i2^e z1J3Fd>kDO9(pJ>lVoxynXI9j>8l!Vinz$ONMagq;HEPaCx-6_*H+Mi9SM@R4Jdi0f zN=3y_pGGr}NN$d=G$Hq$Tkh1 ztVTuVnh-fCgXzF}gz31+P}9%D*E3JDsNT5jeR=pM;A5oUL`5I6;H_Q`DsL((wy8k#LUR|XjYhw|NFRqMeS|z2L_O{c#h}*X*n6?p2S?`WN=;hc zlbMo=+>J0x$f2e%Fgy>`Nf&-%=?ywcjOpBnU=nXutP?yWY<@ zt51}>{wKtLn?s=o`HbjB{gHgcWc8F1%AsoVMcU`QK(piCBxzSD2lK7urwn&}EZuVs$!4-N)%9rJ)2YnFUBpy8G2$~vYjBe&Mux$E2ZzmBF|;CLpBl~+iB}}rM2LF z9S&INWNWG(POoOEnEk!EogfgSU|{w6w&Sx@{nk5|a*@IIydaT(y z(b?i0rON4tEJjqa!Q#iwkv3VnuIx0*q)D-<-H0d`)Rlxib)_= zWt#40%yV5%az!qmysWMr=MFNkxa0E8%4jN8zADYv*x8O5Y8RTCL*^)@y)t|-qr{C# zdtFrzP92`9aTyWJnrAMEcj|+a4@vB<^Vd8+%U0I1Cv!4nIEuX)ZsI2-=fc$2tvp;s zH&i7##jOD(k{Wsk=?1?=Rm?%t5~i+v(*Jf{FK25SEr8UK`DC54+p+pQIdpGL5{{CD z&!LoOudQ2hNHBPq{tnlN@09C@7Lx~RCLhbQc~^J-eqln9_3d@tFuwIIXVx|q6!t(! zyLMl*zn7pVK6of}+IGE4kwSa<{3#su6wVk@<`q@8J9ydInXho8!l%@2zQH_u_u9N> z?PDTj(ysD#6>&n9BWb+#oE#VFf%N5^2F~5~SZn`9Z!0Fb?t9@4{FM441TrkeavE#I zl&^2ya@p+oYG#qGvs(fkc3A^)h z64mk*wGnqcdkue9GArG0C{@P7a5e?ycImY>WX^3-$A=R4MCY{B2Q+#J7K#enlub?V zyY`CK);}YvSfd!Sd{S&b8v4@Zo?!L8`d41MN6!8)CwHSopD4%0W_ zH;9DCF48EpgJHCQ#LWMZSJCmjCF3h^n6Y__7YZpZ4I-{MYz7>94&lLO>KJ zG;x{FH$k4zO?lt8{H;Fkf|y1W#;MNHtCQDW)yk_WxJrF0)Ap)8heMWDpVdNEWuYve zJ|9vs#H{?$xgZt(rw9DBq#z1XkQCqkBnR^B#Y(|lwJq}P$=3M<2?8$NPutNtZ?rz$ z6cb3SltmKDH}D3dgMf2pigR&2kLrbR@zdnmsp~U|LOrh!3_|_4bZu9`7`5kXnX{c zBr1Z$l1E>bN1ycN*ZAe=Fx^K4axX$<{HySyAG{)`oT@b6@K$7^flF{USk{ew@bns? zHg+soX!B~|*utxYOE!F&Kub`B>9K5uhe>3CXR<-#PS`*>O<7%iqKh@W&_cPmC^9B4 zxKc_X!Slk_3>r`p4X8=?Wx|9sh=KqKO3Z&^fE#@?ETUO3Aa)MRyjw@|@mr(49yo~@ zRL4O$rS_hYG)`O@d#RL%z`8w%h*GEq6 z5a_1m1a0YJ6-DJ_5pXCayNagHbcoVDhnE|~ujd=#3U#kpt!~Hk+`FS-(2^YyZLtP$ zup7PWlD1#XWuGoQ|9;#2Eh7pEN#X#Va~WU#tlAz$Q$?^_Y%Qb>XVQirOFBK?xC_^r zt`_goAjfXnU^~r;RMKPcP1NY^H}b1IQ^1|sG1AV*6#OKt>h03D>rFYfS)ikW`lIV* zbL1hqgZL#r(;=3N#a`YtPMAoO#N%OYlPogwI#ITDk>P+*`g_9kds9fzKik2wTR{>P zLvj&_#%+mssikb$yjcH4Nwz85;zIWKDF-5Iv7DRD-@I{uJIgMePpw+p;pbmxiI}9NMENI}#O?=cy&nQcIxsO-NKsVg^%7khtED5}vEYJc@w>;g5Q| z*;u(1Pd`4!>(q$}66^k*wx8$A^=*U2-YuAJH??q)%USRKfMLMb-Zt8bEA+-kDDPyj zzWByUr?FGzfI-@JCBv+1x|8EL&2gEGKj&z@Kskgt5IsN~WfT+L!Yd)>nYWE#Ojnza#CKfgcTE7hNC;$m!sT+7l-c2X{uSci)b;6cXyDtOD?=cZ-M z+{{W`18)dhGfCpAAI0&jF)qOXa?xr^!Tt=zbN-x?kYFo68w1$^hTi#hgjMqGyRL** zYMr3ty3v3F%@9skap9eNh_G<|lKnC`TQPh&N-~!UX`SXJWXQ#A$Q4vMs~)A?UBB3V zgGw%WcMwn^22|dW(@d_(D@#s$PrI>zTa%x=_Q*SM{Sa1+k-x6myy4}S<27MBc?=JB z%wWRie^yGRm#iSePTbnCzHHW1J&^!nl}H$4`*bXqi1XojI6FYH;qlyq|?#U z_o2{n*M1lv|Im6qMpa!!No||)wk)6SE+sPoE6cK@fUcZ??(y?vs^`gG5STc-pZ)ZMM(1mmM!V$JWVI*2!K-9xrc@BC(F@ zyDMpu+3d1ny4zQjv)`TFyNB78FA?49~&!lS(=cA8J}_3cS~=_Ta#&k zbJUIh>+lgjyQ=vM0Xstg|6Sw@8$w#7Cu`(c$K(y?I@RW^wB@NYzKHO^BD!Q=J-LrM z$E!W5N$#mdM@qy;N)>@=_+Bx)(prOhFs3w4#!R);B(>C&2hn)~ftLEHYk7z_>M9)F zN%~=whAQ|k_}ZOn$}X0>SAq()Oe}?iUS2tAggo%x$%LWy4kN1y_XiMM%xF&JnGSB< z6vb%`QYj2lo68Hs%L_?;Z@MQ2r^rhLerLX!CI626@x;ybC!hM)C1g(@C@m#1SVhnc z(Z^2oe308BAX8h_@AY&)_!_q*gSg1c)&o7HXqZ0NQpWcY>E^vS+-<6L$rtk#B_mc? zFVNPN!_!IV#O+83r!sXE6;%|nPMhiO-TL;kXrpi&frrF{^6w%Teqkpca9?FHxNysd5}3ejNLKK&<7Mxd zDwc{Wb~oyCVfg35z(<$dwj$3Po4KZh$ndv?xk&`yE%oH(40WDVE%1)k7F`RjeH)pt zU}k4H#cymli9#p!JgP7}s&HQ$kBb?=)g41c z6;o%afBJM;h!wd6TkiKxW@vd%6Lr1+w4CW>`7Z*nX~^MZQ4A)uxKM)S^`|+a5)w2E zjimwZ(+ZK++}|;%Fej8wR7-)O3B+x9RC8PM7f6A^0L3J6g18D(ST2~lAiT+ zcE2HpLTf54Nupfk+NY3{M5Y5P!h~|Bgz{5Qg{z(l!nDfz6w3N%Lh&mTHs(8>6`2v0 zy0Xu8ko1wZ`boC>OpJ?Nh9&Vu>(^LeqtdL41>i~djeXbiE>|9TFCCvH>rwCT<02K* z^yP=K>DMZ$Mh6L{=6G{xh&~h#oGk8A{2f6Ga? z);<5_cW4!k!`)F{OOq<651CG{_>T(W(L9>jH5JRZ(vQj-JU}Ewd~oD5o%_#p;_;wJ z5FvEb@v?V-*Wr+%m>tN;`+O3)nNr&KipcWpr}m_aPyNYo_iNjNP`(m>*-YL0nYvO0 z8Wl_$6~98bxHapp1GzQ$3Y=*(L#zX1s+zaE<8q8YD~Io+t}lBWe`(Ovdi&}*;X75~ z&>a1w9R2f|;?S95+CEKqrzU*;4pP+(jnA*n=3CyOlKiZ^#6LVdnPedy!8o(q;=s5! z8rxo_B9m6*Q$@4HZSckD;g$QYpWf-Ghf0>nN!nTTqSJ0Z{&C!>lC7uzJELI@tm0ax zouidgry9bQvy=wbx}T~WFUti|;7XHGt6);AP%26#DNOB#l;~5G=>IMa+(eG&K}}%l z_dn`N@ja?wdQ|ZXp9iwHytjAyOT_Zjg#0$9;#xSj!F3};*X#N5DYx-RqUKhX(afxs zaGGVfWw1cx#>}hrRf7-q2nU{= z(Wgcso0iaWY}_x+mA6k0gw)mOJqecu-uZy96FI<*{G6chUAA4@|&meS(8 zz=>SYgO{D%t0K7bCt|Q!OvfM-J`#24iw+YzyTyJL zkQW(9IT}g*Fd11-JC|etW6D3<8R*qtgrDU0hfX)%%QMWiz$yohbw^|spYAW_G`3nS zn&p<@S?6>!PU}-l>$3=`!#iEkgEOU}Go{%s;dFiR;z*B9{CAqdem|0~EaV|}if|WxlG4oy8V}~kI z<~y{21g>N!I#>xZ(P18Ui_L3VBgw&Cy-M=i^<`bKX-2=HX;fc{;|n0 zKLt&c&Pb%HcscNbwn;gKk*FZ@!{S?uWu0Ih{0`e`U7u+v?Z>gTPaqnHdAh$4X@;z>78*$ z?nH*{yH3hW8OxmR&x&3-HN861!Pa+;R72^UIM3LKt&0IVN*feX-t@8+#~^s$F@0mm=* zmh*=}mB&w=q|e8B)y`fcY3&3}J$~!+YH`=DNQ8ybIdkUOKVTTo)8{{G&?gnBnU{{M zGN*3Ol&|B|6QD(L^k-w;0 zWmYh0qqG~4@2ra~G8v<&s{(j?Xywv{^_BDB_n!*ulM5S!jv!QYDxX|$ka^#t89RD! zPwd{A8&Opx{Y%6{GRYivWzmlYIcC!A4%%s37Y)&z_M;xPt;y~!0Vn6?DesE-Opu|X z>?Zyy&N3q)t}QRpMlx@rn#l9-2%=A;Uu9$*iQbJSEf~i8Vn7o!O2mZ5Z44ietH`Ji zaIjpt+7!Vk*JsUZ zQg@_VLQ-R;d#0p_xB>BxA%SWDNK{A?po!V*&{{DjhD9w9Qj-W#9M|3Tg_0PpCF%r&De&I?dW%C z*1eY{PMMo%kLxR$ZH?qFQj3@sjO0no59rR4tZoKX7NL^VZqAnLCo;;ZeH>oUX6Y97#d~a~N_XCkT6tPCC!PZbv0M0s(?t=4fqmidD*Vdb7 zZS{$hF0+-$sk^!s*xEH-(j6;$P^M$ywbfa5I;AAO_*3FxBC68CsO&XL1&yM?C?`V7 z;qt^x3#SMX{z<9mD8IZd<+;`myH~^=PNSLBXb*ITj`^vrRJV16hnR2|c>?1OAn+iUe>V-9`fG!jx`FumPFLrJL4qNFYHq*j-vt1x%G zt4qXz5yMXO&SZmV-`ixV{y-lIsMHR%F5Q6N^XfG)VRE27dR$7mq``8bQA+t)gM~!! zAXfZ#XO2=cXpE&ZLCI~i{0lkr^%h1peyw5_>x~C<1 z{537{wo0554Zj-sfhckdJ2lmeWp(Ipe zA>lkAC1Wz7J?I+Ry6rsuI1xU&w%b}-;@chLhOZywaN`y`>53SVy@@FcU)a3&*Xhfg zj{6qE*gf>`va8dhk`+$YFPuhycyL&Lxj}2I%@QLuOVxl`_tEX0DWj*5h?iJvKup_c zT?rW`D(2l{)LnO=5XcHc2T?YiP4zwAA1Reuo73_hv&~LwJtr}m4E^e~M4qRF3p{NP7fECNlNG=#y9Ca7OA*wqU4`gTi2L>fMo;$!iQXY)}@JCrapA zN=?2&TCcmmp!Y`v3#l^QuJMHhO&~x;I{WjHmDN;EqQ;?#%X{`220hW$#OriZvi*X) ztCB38eo8E214Lf=$y3?CM;eMxhTw#_ZTfMfff9k3BGoHxUo04BGlzOR7Qb0hKkJ)F zeoip|buC0NM=6Oi{+G4F99C5pyS>HCO6!wU?mETt{8_4_&~vJwlf8oTAF~z|Mm0XS zMO8=5m+^;yWEpze`w~LCW%4X4C?!s2!WDNwut3lQ--|y zOo;+7{uhC4Q)}`p`b$CmXx|ZcS&i%pRaqHvkoIF z>^_t`x7TcE7e3eQ-$z8&B@2$So)mm5E*;r4uW5029-czqtf9SvKiJ7g5zoYgI}(MI zgbGD%_?IR)ePr8fjxkJwj{G}PiGnk}aEqz9?X&|;%BogWruRlVRLH3TqN&CecT#n2 zx0uqrx!Iv7$ofk1P`SLCX3=azu_$+ix^7NYq^TWLXPLOM?5Bv^P0zIk zh+_7qn|#+_n>_NK4i4#bQZQ;n_Jnm?{Ma)8?w(9TN^)QL%ITJqUES!7rn&le?I}p1 z)FXksY(z%Ak~W&mt%GN*IR7`R$hK!1naT^Lb3z0m@knQZY&=s4ZVs2`puSuLv+#AA zo5C=jefl<1(2UCI&?m7#y}M_rt5cm6Hu{uT>EMHE^$!8=43U81p5=c8pp=H8Dcm`2LS9_$C&3sT;b-m`DsDQD)lHLyV%3P)BU zHVbYFN0uWr0CX@dmAt*M^j@5_b!S?~?-EXW%D~S^3Xb>_vGv8;h|ftv@xO0t<^)3e zZ{+jaYTArZM@h&;IP4Wm<4ra6Meu9OOo}9YDspGH$_g84zMCEpzcRGHfUt>Nh&Z+jA%j6RsjoL(ir?SsXH`>y?85(MS2*UEfsxXThI{nvwik~Cu9^W^N+ z+K(1p8+x71dho{Wwcg&BaoiU$GA8tap42xw%R0_0)NP{5I#_fwI+C>GF-Q7yK~vB| zv?sa2YZyt+^y7gFFh*6C^xfFo0p#ocRgCG-kd&-V#I@EMPsS&%{iAErG+p~|V+gv& zp7PecuH}5kAhzDLDN?yRU4W?!)yg2UuJJ^@Gb|15H{|Ppa`6$aA(CJ9+X~vPcl9ty z;J(P=wwlDAsD2oB#yIYd##i>sjtKu!Hn)jt>^LuXCe;~#3c+GC?Ia_soRK_%MO0C- z9YJQ^+*ZfEA0zDFeQL#rC?dSf$94{c+80-osp`_c?PH45>Nb8w9m=D4{@o{EqBvh; ze6A7W?W9c46_w=sX4p5K-qV6P%zXyQb8esScD?CzNssotz%On+)T`hx)?K8mcwiG% z(t&Ez{l_#LBS>G3AdRJsJqVabjiq%x2%JWq^3wXCo}9iww!7AOH%e1z#E+6q{&=GR zSGrz0rYhqi@>}-$mx>R!7>B=?3cStp@cwVay#-KQP1`N{CLwt6z~JsK!6gJIxVsG= zEVxf_8!Xu1fx#UH3GNWw9fA+;?uYNYr~Y&Px^=7Wu{u@H)LN^jdhf1hSMAxm`Zb{{EAG{#Yg8#S?z22Ib7o>3UiO{F>oh=tFqr-)0?I>Nng_etlw4;3~ z0mbXP#Rkj5iTL&?%536NB+S1QwZx_Lge~H=e1Fl7j!>M^rJjMdXv1YS#Z&lh=w7S8 zH~oC117Bw&w~zj;I?$JblvI^Bl5MI>n8O{UZRmyusXU)RK+;0LVb`8yx+vVI9Tx7F z@>mfzOZedw%MAHx#5Bbt@k4hzXE12fDR}G4J)bDa6S2fSABG?Gr|^ru^&lx*tmEiz znaF91Ug8s}v+LlE=IhG#%StTJo@)rj7_o=2qt)w+3aYSOeA+~3OS>yX3`B@moHdp@ z6^5c$L{eb$0{5Nd`OLEk4*%W#T=9Gm-YVO$1J3pAwTp@1j%91fHrsIX^_p@&Ws?A#(}fy zPc`mimIM7iaY0_@5l|Ue84Z3G)Ak;tI_($31}nFQ zhoSjYjEqW0tqLq-^t;vh?3ZZD7h;dcN@Hjq*HIV9un;ZoCED+96>#4E{c#OOiTu9* zEFt5AxG4YMD9pe49HOO?Shz9t&0D=oc*U%wrhJd zM)G##4h`3K&)Tn=X_ZFzjFDO0g=NrY!20d_z8sSF>k5q(vo-jA)+_^URtC0w8Tm#M zWtCk4iVy#{1xOaWf$qT3t!o%Bn7yW#2c$pS-)JY(d>>`JPkyR-6Z-T8+IhTENaMK? ztD{U|CPEmG;rW*EU|=0^fMQ-o|J2eE6?Hth%AMk>o{NG|LvMvyE{Pj%qg>B%M4uxW zH~TumZ#Rj%DJXT{1sAVDj{Q4h#mH=7w;YU zoB5QyaI8;|BB#zVpqN8JCVJN2O^=0{eQU*@^;yuj3$suq4kP<=?=-*7>k0~!=~I2~ zAbO8-F%X0e$CZUBlSOc3f9~?#kcT+hOgA@6wY{1+q>GDp_)@u}B(}4H)G`pT8Fa5< zb>QB;Da9saaAAc@;_P=d6Rv;o_MEe(>GfQFbg>iN`oD>%uL^wqNHa|T9DIj>1IGn* z`M!NU&-zbBWeO8f>=&CnRGcH+GZVGngU)pB}n7!EAyBxB4oSRX1)r%s*6fl>`-L?@g~F#L7eis_uKl zLe3Ft_7l3R_i_J}D0`RP!?=tIrY_C0SSn8B0wHXNM0ZS#6;m^!#=n`eM5^k=ll{+4 zSe_7SE)!(T_UPmCvwzoIo~%Ayu2ABSQYwq>d3pm6J~QuNm1MhuRGz%zn<~<^`L!Hv ze5*zl#C%#HZg+rYyNil@#Nzu1F-@@mjVE)2f5VoO#uI|tRf4Ey`wOa|hh&2tUCsz; zpFLr1Y{aJMIzsM20+P%eJ(gj%u$-&k+li|Iqh0TK1N!X?f~a$Ky$o%s?}{0lt^*V` zqI+>ygI_oXK?sbtc$G$(Z*5r#@8rv#P*ohYmP!{57RA`FkfO5|N;WkG_jR2Nx4!Kx z%j3!xC_fHdqR>pyKid@3V<}{-m^3X>1Pb5hWk-7(7W7Kh>r`kXILKRnz40QpO(ibO zRBu{MNmrgTE}wEddRV#NydHX~&N4gzsP${!sN#-1gT$|*vuKX4r1(CBv^?jH9W1Ro z*rgu(EUT*l8o2DmP)vi1p3T`c+VQ1JKkUHq)~EMu#W_He0e7040geZGB+|ho?l+%h ztmfqE8ML+bE6(qPU(A)&ma-C8d2I5)wpj=T%L_b1bNNq&r8r-n{Wr!|l_@}CXUbV3 z0k==Vpste@=?{6w^b@$MU7GRvw70?}gR7?PuKEL&w>?$2#s`xiVv ziOB42Z&mU={@giM;lMpVgjtc&`DD5k@jSM8gR0u6rp9MhmR!?@ca#V|kU2tJ`-yFd z#Vu$~>%Pc)qf;*rTIwyZlqm{DIBw`a#V87Z(W>}Q;ya*dxG%3CPgm0okgMt2ST=3 zg(!b3{Z&e<$sx4^2uK)6MvtlbEN|j*Z#MKLE(yL?#A=t9#5z$jSF%-@K5bWyAWblW zQ!v|Vs|9LenjIzlLw9P>+w)@3i}3m)c)Mei-2wcFyRQ?n^e3Q@kuE_(y7nd3_4qrn z>|O?muHd>{xzS*;NjdS)tY5As zZ&5TNuuQE>U;q7JvxVEQUT?{<;apqlez$pnxffpYkxG*7hGV+?b~nCOgx?OyYHLlp z>?jrK`8>ZJ+`-6mYC+Y<5w{vv!or0kQa37NPo{ zt^|&{qZb|=G|chchU=916WnqZ(A9a%Qb=&2h3!Kbm80}0<%4e?f|kF%y=S;**)VDs ziiiwM60qhc4R$_GH^FKL@5v*I)AKcFUsM!USjuars{#FQ5M#Ti%U^cp$(M|OWf%2z z;fh?@ed?n{a72euXDq;MNZh0%))kX07 zLroHFyz4^vs`3<1&-2xNU{>9FJvFOo;rkaf%!}#r27N7MDMp!)8tpq*!@mz7n(;Kf z!s9XE5rcxlPa3P2n|}m_Pe2;_e%>`0V-kC>8M$h>vi8)k{4QwACj3Uzqif$ANDTl6 zoeujdoJ!QB$Y7 zrW<<+F0?M*B=MsjyH8>9&pTIYv%- zS|7IW{Q2$Wg2`^A-)2;YjBU58nBVa!vrTz<;m}-rXkOWf_#?B%ti-eINMyS;*9zWe z55A6WrW(Zd%^-ep5uG)mjaFh`+UDaP`InM@DIXknEYVaD_nRZw)>Wx>^d7ld(`v_? z>33OTla6;qVFM(V2{%oKIG$L=$9*cF!0{b(sO9)85qWb@r1n(&dVYhY2x9gt`%O6E z&-@eJYeNK%#(pzv(hFw4#0M^=$MiCL-;(K-fWj(uI$XZGHD{T|q}D!!Ll4 zkoY$_;qjkH8zO7pv7eF?vZhIuqWD&}ott@1mt|LyOCaWa zy@qiF-ho`k7WL64p6#LUk5fJ%`KjUh(0AS(Uas)4tBDw%@*L*4-v-btzLrG)d@SIVU{-jUmOJNU{dw@=`+n)*wzpog)zP{9sbyMKyj66|^#~ve>KtXL zwpB!=w*T4G+cZVMyScvFFm^s#!`$3rUB8nCs@Q6+fb2~uikANLnIV!XNxMiU?e6EC zUqH?`yA?(x{!SLOegQai3SqP5&Qq&??gZc0UnZf#*@_adw2a`-j=%Sax@7=5FUZ~S zTUg+PurtAvjDt{GQxSYOrug@MMZA39C>dpm+3l~lO9#Kc(CfO(>(0>LET@^>j=wzs zk?Iw(yw|547qOebnaMN;u$(|>iUw>%JdW9~b@=VR)%`{-F3@CeR&I{M5$)=j2dH*g z0p`QG(b2O_O%9!Kc|tL{tNxlI%B**ohys;x*~i|KgzowFA)Y%uzOpwy^#msElpk8< z#t-C$R@gqkj|f+B!m_G6R*6_Kk=sG55I=kN6~t@r%l#b<98-y%ZQ)PMVguX4%#M?~ ztq6zMT_$+hXPmYHy1u-Dy;hK3tH&R&Rs_C%tut?_;b5_FJ(IzU18P~ zmaLzoUS3*26fwT2!uod?D~}3yXp7fg&utD_A)cF4nwmp;RN7<<9k``HqC(A~JQ~~} zZDN!ffi!^(X5&5^Ckn_huO~=cw&~fhwV!1y<*IWaK#&+vxW>z^$MQ`7brn$3J6;YN zRvd-pXE|x?{c)Te35lK)s|USRb3{(%8hQEyjb!|3azVf!h0vF>fb5%U>YM6zTN8F| z;-5-dK__7!iXS2R=TL=Oys^T;q$ovBPYt8a)`zR4=F5tUUewm-C^8R1f~gKWcB0?$ z#V4gi*p>g9e0To)XbM2GB*idTLY5}X$uE(4ci?H$Dj(FeD-~DQ7YLe^RGv{g{MF?X z^3R-g6$+4Ltq4jg(*5012T%}@N`zZwLIR2I2q5eWSXdvVpSHASCBjqDSv8vWIIf}% zKOh@do3^qm#s2AUG>Tg;>1M%d-x$x3y{b4QOWw>n`RyrcNcv07(q*oK;`un*Q*FUi zPfdzUBKGhi7yaz@!y{K!<7@~ zrvj_qQdtubrc>+hy7l;SpK;@FpFYfrZPq>$j>-p3=xN$DjT_|G9H2o#Mrs*CJ(a zT7^Xqkj9p}*f>`#GU=Dkt}FF(H#`@Wq~LCYso99Yp4KTgYpM$?78?g*_?gvXY|D9YopkiL6I<9&uN?HWbplVhqbDY99fI@S2^yA9hNil#GIyP_T_|H4 z^L=GD=m#BpGth^6ENuP230pUu4f7x%7+DeI-wa(aM4UHt7P$%X`_bYr%F3RAgptZ0 z>d=Vm`XH!p-f*x)@CknGD$XBj=Y_N|`|tYSiM3WRoNwrk+NP$^dsY6Q!;^vG$(R(2cHev6F1hi^ zF05JA+6iNvk8v^#dHkrsY}M}J6dTICUC<+LvpPOtCIP}S>?6i*hJh$yK)qy-TD#_8 z7^T=^-U=OR?K5PqIh!2v(%Mr0PvV5?`E&#a-fIhQ!8L%T6oKV|f!$AU|FM7cfCda{uOTj}c zda$LsE)$NEjy0zI3kKX;MSA$AihRw zK8qE!y`AAoW-27BFI5Yj3oLo=mI%;t?%}**4>Y+QqyZ3 zbh#RY&KC~{l`8eL9IHqSv#fpKo7x_5^~FpGfnxlo8H2u|;0S1Ni7?(JnL+{k0$U(N zAT<0AIsDba5>i^uP`EVUto5%iM5PQG1Di)4vg< z;t@!s{U#cPS$_61bkW~9n&Wxpgu7WXgV?&NsgxHiyZ>drP;|I|{^nfq(L! z5-a>i*_t**fXcq#(G4z6QQo^idf`ee*Xh1CC3L$fpugq&E^NoCVFH#lUj61C1%b24 zGHJk=bvvbw1?kGF?X@Ak6C>Te`vx~NXp=nAl5(noz!s$z?o}$Ddp>v20B6Ps6q_k* z!A>-A1P&R2BcvO8v*!c(({I-7e0c3D(F@$fj^!jgll+vdDr}GiEZYQ3+Q@BMNfS1N zy}#IkLMfe(5#rlzCIhstI#;-VkjL)rrrLCsByS}<6II08x`HX^|a~dT!wVL zOLkaq=F`Z#CZkuCopiy1b0C?f(F}XzZ`*U*kRLu!j|%Eb+23Y8-EqdROc6 z7Zks!vjX1_d9~Tn2k8k<4wG2ic`f+DAL%%SLT%!c78?Z6P!WIMA^%7R{z#8~^IPCm zym{jg^Tt8+3?FaU{DOK%ovf4>unSja=&b1|Yj84aU9BIoUi~O_{CkX@6HQ!99%c_wHn?6Sw zZX4uB2e_UBaFqOZcdygraH$6r{89RIFz1Y!BD=N^%VlGqv_9Ch z_Pu3oFS4}$tH1)gH>7PgnbL6Nv$nuU(S{A8L|qi6iT#Ux5DE(42faOC`!@^06^L#q-;X@6$%~^ArMm zWxL-VnU6`49yp-neP~H9a*^JvAmO;G?og>aRq|uupklpN@;kf`vff*(btO?I=a^v0 z2!}f?w{4&J>(j^dlOcqrh5YJSXT@w88K0!s9SAoC4O8v9WsNB$Z%;k$PgBU&%A?=( ztY#(G$HLZz8bFWC%t(gC!^a{|X|OLQh?8M)>ElFdQ%m+Fh_EEkFQY>Cf5HqgpKMX7W?GNkB1hLbX5hqw=Ya$*B(3S~ZGL z#QiW2Lt}I;fg|~_wdyBYpHycTk+|i->MI6oE4h~6q~P0(zZXW_oj;tUksTa zqadLvP)Ef`^I8TD#;}L5aD=cF1rHKyfCk+xGGrHx$_MAkt^-TtiS3h*rRsd_F;gtJ z!{Bk!enTrESP6g36tpY?RLwc5N^3f3YWOK4SU4hB?rF}$YIR)qrqo#-BN&Zq31AIO zlb2tZ&HTbUt3UW#5*OF!T$G|cy&-j5g%*+hr@QUtiCbt5~B6#K;3Hl`T4fP6)94ALKhRP)FqfI)4U0~m#2p6qrAD+ zDqdNbZCXSd{(6f5Gm#*jQFMmQWhUlGVZjiY7Is%o4m%{3>OMi7Nw)q6*pH|7@ zEPQLXz4^@xbJ0(QboJe}S@z7H^jA!p5(uLTPWCr*cFjWuVkP2@ql*$BONIYv+5h!* znLWIU9VH5`e!bX2_y8e%Y8+iO1=6T@YW;83`@7Nw@gaivAeli#R*tVMpLusTZQde= zhTw8Mwuz!V8}&okudw7Sk0T)O;bBt!n0(1^flL6asN{$kdR{ua=N|hF%UMFY*lr}| zsaE$op(kO*sRd~xsTZ)tTVRhFmnCYLfEg#@t9u68;} zipult2v+!_NN-A9>I}=ikXn8G31CkKo|;9RnF+jQPkt3h!>p{FMl7vo|7!E1Y0)uj z(a92V%An#@V97Mu5Gx!Ab55-kzFDLR4YvRxVK+z%H!RXvhFSDuw3lcHH0*F>*FOE) z|D3cRxa`5b)7pT&0bdhOo!#khNspY-CE^fdNqs?=HgndAp$Ma;3)@z5o_CWqb<(+> z65_iOKS)Y^iiA&->~tQS@uPn1gKT4~$&lMME1Kc!iRLIw;=sn8#qhn8$_69Zy`Dj4 zt%JrbepmM)OG#bmwTA%fDfhJ{epWxc8#`8>vUtV*w3ULwk4QL|{BgMIo`({ox68Af z;|&9bdX{o^uqC=dAu%1I0NvG)7SMjc93nH5hXLJC(mbyhFZPCo^`y1tKtU95OWJ2xpxE zgmlu>lZcyrRWSP+*mJD~KsWokEh5wffE@ydp>vU(Wh@>YgZ&Brfa!;xPPeiU9yzmD z;d5T`AYYI4k5P}%rTk9cn#0?zQRy|u;A8|UiSxpAjWG0HEiH4j-x{&sdzbfOExnI_ zBA{Iu42Ru)Ynq21fSznte3-G12Y(Eh3xp~pLlly!|CmP%n0tNZ(lz7Ky&l+F=(|HT z!vQ)#FM{81cGSeatNi&@M4V~Sv-|NT_oah`a=+tu<#P0cs4Ph!qj4ko_7ZLmIqXjO zYo!6VwjB6+5aWn+hQC~pLjQX{wcGi+wAVeRT%ER&1`NWn9`WgONIi{oissRiiGI8eT>kBnM($kOL|yu61n>4Yn|9` zRH`w-K>3neH&!Ld7B(G$Z4jrlO?iY(xlK(fCcgbiC@1rEgZjzp_7 z!Zb7IG;{wJ4k#J2pLF)7P~;UvN<1oQ7q17nMz<9`Zu-mT%)aPcqpV9*{EHm_;q?_} z>`JU!j>S2aQ%#xL5LT|NAfd(pBM%uxuCt+qG&M+` z`MkQmwNI9vg>)|U&L?FKk-mO@xae7QE5WdNlg}WXm_OVsoa8Jrr=2(#oz6ldRhuSV z3RRb*tSp@C%=ALJX{_<{MMV1-Q`VWgQ%#uI6%635lBYNWHR}CA)f4Y`!+SW4!~0BJ zLQd)>ws}%&qImJlkRrqJDCdTKojPK%JYW7X5FVZ9OlrFV1NvzR44X5e3!!pR3G0v- zf(Pnz*Th_+Wme6(_&WrJ7pXu{wBFG{AfGh5$FeBqMr~&HCz3y7Th;niFf*#5nlp-K z$(n+7j>51Ls4Sw;K*o-h0@*qm5JoL?vyBoKZ{XNdcwU!h_5SIGn1%(m!iNbjqwRwd zvY0!=aT+RK+Zj9N0vLsH<)x4f$~h|C9B6T94}*ScW)Y93m1SxJm?Ae!!{|E84kj$c zhE#V!s?C6xXfAz`_IoKef&bF`z`lDMHZ3~dg~~FcCdkp9sSIiNpQ?T8ZXY|_SjrzC z>3)3t6Y49R(*>z??(_9Po{8O)c|v);cbyqa6?~@KVmht=T8dv-Xt88N?QI6Kcw4E6 zGATaej@Z4PSf(anmD%;VYJ=Jp^`av(c7hQdq4eF`PSkgugT%;M0A#J0&!-HZIv*>& zt)I0`pS1-p+KFw?*$j0aKk2VDB2fXJHO;%WDbvj)G zZJb*wq!qTV>8N(}n(s%r?+jp6`&E#}U1&ls zH2p>!nvQSa6p#8zrOdjCN4jxA{lbv9T0z0X_Y0+Dtx-MZl7by_^dUATSK|?>AL#;; zxKUb)6%Lk`@_xHC^Folc%=O@U@hyrIU#3+%NqZY;1(y(vyALd)_CUTOr9T_y_R^HY z@kZUMk#Lby12K=IWFUB2Kv~leInz&pcv`Oli6-PDT}YA#k!47uv6~|i`~m4)skBy2 zm}k0_W4bgW=dU;=WXY}WQM2;V+Y(6=a;q7s7L&E@KH-&UuP6W67Cz#(t%$=x>B61v zXxT7i-h?Vkc<%}X!F4p5M{NU(O+HbHz`9hb)PPjP3hztbsdb{Nca~&ld5vk>I8dd( z7TT5V{>@vs8nL}0m`?rnS0p3?_(?Wnx!XPFC#r@ijc#!mK0BD2*@|QVZktNiJB*p9 z3X#!j^&N=DRyL3^rOG|%o(rz_wBg8WmfyIP#Ryw70S#&7>@ zKbtusp_#=$gd_b^lmAOHM|@cqXH9Ut`A3|ZE&c+Ih_ZW{8tyV6dEmKjlJMV8Aw|&7 zrWv|d-{B89L8BMnnv31n$&F0Cb&r0f#1^VSw6*XmGuSI!65ubq#^$BVH2-+}|F!s9 z_o>Yoe<5k}9)ymVtck>ObEwhV^UnBdt)x^BLRmxCR6mLjikUtklbSweVQ4_p!6wL3 z-c&_5`j31WM`4+?(;%wLVB}h2>%|^}0!yjC-i8P_dO9e*f7eXF@TlEoEK`C8ELkdw z)2bbrAz}581|U+aIcFDXBnZkuTFxMtHpz6+o1!bg3gUXgUNXkS?qsRm*RlS_1Xw&* zWf}flYUHq45Il!CE=-)}h#wVGqzmxWB0W9C-Za_Jjg|$Lah#SRuPqqYa-6+ZnOyj( zxyJux)Aj(?Ay18{0ivdQebS0Be>jMe4gwr+$Ghfd@)|?&N>&*j8`ac-i}7zUfWoq z@PI1yZ)22nxYyyS9fOkL)+FQts~Br)etn^p5~Mx}5cRLZGs6)gN%$(z!{E+OmEs&O zcNq64|B#d=1Tu&*Mkc6(%IQQJNb|bE%4ll;tr|;#mInoK~t|=`|yD@;vFAq5nbbW8b3*IUslUBm(Uax2pVZNY_RFT znJ*g5jS?hh1DY^>6ePo_!*{W8CvpJTwI&pTG(W88-e+M2jhm7O|6Rm@2I6YOtUO|; ziBVn1*gZpZ%$ou{3zjsQW#5f-)>xzp^e8k*REXn-mP=$fFYF9Xolh5*G_Qj|k~Uo$ zZnCN~!8~qndeQU=;LYaV>IV@^7M-8n=HJY;p5}8A0DJ=4#r2xP&4*xzB{33Z-ZQem zrOX<}C@8!7HD`KxGwQKJ4!Ga(7p-DU9M7<3n6m;+Q0IQ!Kc!F9gbG?=f6d*(2liFS zM-1Ee1g-n_Po5;e&a@C~27~ICI9pC>Nuq{Xz|;bfDN)jldkZ6T&o! zS&;xqyya`2Tx2;$p1lM2;7)ui3X1AkM-Ley7jTShp;A1{?C#bDtL!NGs9E6WQOeMa zZV}3zAr?87Q^o&}NlYn`6M1KxcK- zMdQjJ7>vP*LMjae(a8WpFfb#&guOy*0$&(J0|P1{^$Ea_0S$|><)eOot5DLp1I`Z4dE%IVd?a4G z(aax|5<^ZuJOo!hFb-W?7ql7&oeaaArA`htrqZBFHXH~fO4pf_JnbVS64y;@wRR$i z2gIhJS-!=Ds})g`;|fY_xGxlzqhz1)bX_O4uPMpVmbCB5RyQAr&8r()x6PuD-RBF% zMoFPoxRa)4SVdNE!hqzur32a$fHf{&MO>0%&rC1a7 ze)Ue%#srr7q2&1cdH3tyau|?A|CuH+t>-7|#Fv4v#$XKm8w}<*BN zdeSej2+a+Q{ycrx2=-K9PS6x#x%lcKAM0MqonT&NP@wLIbR||XCwL99ikb$_h7hjH zf>jC7QDmqo){!N;;sK>>5UiLUbuEJ7Q)x4O!{;n~FIsl$JVff&cu$a-T2RDlIF&z2 z@c_Y8nZnhsGgyRUjo9tS$?sX4fxU^A5w%)X5|N1PPmxo)N2M{(Co-(G+}xQt+x>z9=6po)r~+L*4mAC!=moCScZu&2gF51ce|DI#@&8m-FyY`OQJp`n?`WBB zd;ME`BXDO{>hD3w6$^wokP`Yhax<2^|+ zCsh#Vnl%C<$!p%!fKrIi_IU5X*S2+8QaN*uPb9_kbX>3Y_zXp!`sV)fl}EZ;+ELD$ zWrfFPKTtM`sKbewJNdb5;lUt zO07cw$1^vm=COS6-8usaR1!cDBmGYgHexu`sSL-|njba02FDbSA2l{l67W8zi@rb- zf)m3a@2Ma07oKw9ye+!k$>TIV(E%kYv&F=> zjy|og$R-4f8NAGo@of{_(}Enh?cut}a}U5rKT=VF6gXsnlO!?#(-`SIff$ZXl3rjC zr2?zWnL0)&2Y|F!MvVe2By)zVP5~B{!8>m;GB1r1G+enboy+Dc7GA0N7@7se@Ftt! zDgQ!x;F~pEh-mMykIH8wZvRO|NJSVt8Oa%H7{k#_;&xc$*j#Ilby~xR9<;Ds)Xye_ zVLuE)lJOJu6f3eX#P;S8Lqs3}BJ&qetQk^Y`@mfp&Cs9wNN*2j7zE@uxq0NC4 zi(8=qJeS!{HwoMWV9jR`yjIRok4+eMH*epigRlE<0U0pk-`io2`94Tw^ccn`J0N;X zU+kNfK7eO5aAMz~2^7Gce-@(v>nHAoFh?_1xXv;7KV01PMu}xM8)BlT z_K}*~V%*%40Jh3Uq=fmJ6wi9cSEw?l}cTc=w3K!82) zU1uLVo#1A_>>&}*j=P|0By*E5lg_1Kz!ZEp@_Uo7mAF8r@-wud${mBxlQWz!_6;Wx z-K@`Sc&(SlvPIGC9@7ZjqOZ&6)XaBB=70JN+OH14olzErbm z?;RdBug=QFj~>d2H!ufv{`vh5L`^Kzx%8_Z?=~%0re2a-$c8mLxo-Q}wXqrBXp`h3 z?hHW2)hBE&Oc!DU4*26q9DYf@1BnK=azpw62H=Oa+?&R605I}9xw>z-{5~ff3m7RO zx%q&(KqWLb#S9BdWLs7;?*qt!g(xO;ko!`9$<%(KUewEQ$H(OU$>|CL#0Fph(MkJk zhh_VwO%-U!uMA6LIDY_`PvXIQ9Xh1uG(da~EliQ0oQDMJRpdkiVrA@88~w}JArYLr z1RlS#?S)S5GRRsfiCUA&8wa9X;=V^=w|b)I5&$LHG5W#!!@Ipp;H-9}zTZY9-m&__ zlfCQWeXMjl8>cz_tkR%OGt_8_&Fk2^N8a|)sK?Fbr81q}7l7xgd6iQ$4*(J+WKzM6 zKRp}B3E#~j&BtBDb?eBLJ;OJkRIJbH^ zVea72>vGOC9%*&pB?mQ;gvA-4h^0?OSD2375?oy4na*_Ku?pabYMnCcgFHuW27#@I zy?f6jVy)3#dT+W6OeEXoUd;d;>E=X%#4y7%1n|*l))+^Bk2ZQlnL5>hyOcP-1{Cv`ZoLYZ6MFM_7NSAp(PL^?c$6u6*b=v6(iKq-BR05eN~g^^7_aMPDTLlO4iUGKSpl~^SXCSKF-H!FvBcSJe;+~$3U zo(|Ww7(2h{0VL_*H~n<~4M?y^(oLM~Ha^5Ca`5w-A`y-Ph|;UV$VLHx^kp>gRj7+P zpwFw-J20YJP6qsGq`5LonS+lvV0qp#x-p0~LYYH=mr&xhvjFJ7BCi020B`5kd|tL= zRU@Sj-mYfMP9+_LJN+~W-D4P{cq@VFF`S}!yglUIy+W>-YSLpjDKRXK#fUqcrbzM4 z9UaReFaS0HShM@@;7tGw1#y(_Z^r_{J1gApajIZbScK=UDp(vAfo6TpJEpw}+He|9 zRTLZ@hM@EzZUVT|KM$92gvNf12M*eCg>t2v{0BIH9<&Q&ghlXK@OCCHY*8dGAgEg= zoF%6EA8}=`d@!D6r%Sn*ZqaGKZh()l55GC6Ou3-Z_dQcQO*v*BGUxm{xTE{LjNhi0 zG&VG|BJB4(`dCc92hw!9*6p#fvXTR2KKU#;T&;p;M?mrZN`<^p_0J;*z4Ga6Bl>#O zI2S`#-(IZ6vMaTRuJ*em4ezh+5B5UZumrbCz>gQki_aic_<>nkJ?Tpq-fr=^pYL7L zQ|(@?{UiDH#gwb4BXPT8Bqr=p-7hEn+3wTJL(lT4RjDr%h@CC#>EWbZsp;TR>ZW1! ziS0IF09Cz0fh|na*^t(4JoYxM1HfL){#2|G)UNR~mp%3LvXsJCI#crqKd7{7OrBIS zO*lNbCU0$D$bNdXbQ5|poEz4r6YJ?`dBOuU8GgHPJgA++PB0xRh%2hz7D~@87_cXM zG$R6)6Jz>=hS{R>e)}k0w>*{8y&J78xH3L~J5Ju!pdX*RIVFu~-&Yr>8L#a2%ZRB@ z5nmXo%5A*dlfS^w&xc-Q?cFPPI&TX88QIx#7f^1`8O}m{xv!L1yiW%`&8>}jo#mj5 zT9l@*`O2PY4sXaVP<{*M=|e z%u8=#qKco@MCj_nDt`p!6E#@~={>7r?!kdK$V!=SHX61T22>{NQP=HT>=#cg0v5cp z6&S(qa$Bu<@jzJSJE8Z8LoV6yuMp7UnVl2yDZ6Z z@QT%dAnXxrDfJ9TA*>E6mfN%R-rMxD>2rHjtosS*4_{mO+j_KdybvNzhEA``e4%D7Mc|AK@0i;{p*6JnF9!+lgb(x_?tPf01?02tn=3G*Pw@^^ zF@VT8%&cve$kY;M;J$R&<_4vc?w9TQrRmYys<=x^Z6p zZ!pg+)>)zlsM=>(jfN*R<I>{IPu7&K=)_b~ zV(gjK?LG3&g!28*_sA43zF?xQ)kSJB512J=?iV(s><#x=EIA%IS@S+k^?gI!VmV@` zN|Ud^9c;AI9HkYj?)Eqcy|MoMaA+qE5|L~tY-p*mF*SrnVp=jOKhhmRxgKU7zM!gi zNQemtRn&YiC}I0Mah&2?@;QHFJN1oBQ`9^ANe{ z#;2=#2!o&1t3z)Dt*BH>=zT5%~~4K%Yn;GNCZVhRRrJdEG4sSvQx_Z~J3AN<>YeALK1bWbH<_BNUM z8+dz6f1kFDi*WOe-+{N0`rPlV@Zu<53hzf=_7B};19oS@(@!T>Xh(jvqLsT!OZf`0 z3Yk@Nk@Lv|-~LUT6GuZruS5bObQFB+(l3ME@X%Xl4mqpQaVze3e%qCs#hHGAOFIb5`bVnMa zS&uIWk^Pc}JJYc`_dXj$P+-X2UMf1_;Ltj^fA&zR%@3Cr1QVBfHyk^~N~OT^yJ@-c z^v{`>_3KK3e}fC`OQR6eQ_>nkp7~Iplmp9kUTAiaZs^5{NBBlS81FU{exUD{it?`C z$yI%f%|__{=XF=Y$Z%rr)`KcSSQdh7T#49%R6f19hY;_^=zWO)uj2AqhpJ{=L$DVP z`j?)B=ScafRbG>CMgVDR`c4j-DGv{7#8RJ`%loSI4dRN}#m;YGgQrSO=mg;O`tJOO z0<+`sKKTK;&+I|VAhFzIV$p{Wsn6Us0^z*mr|&*6aro}aaR|gMrrJ&IIA}Ei6OY7- ziq5KB{;@|-{)@2Rlb1R*XXQgAi4N^2q`Nvwz{aN#G)3Dv z@veV+vTK^x&>U#Ju=^#I(#v|`%s}@*zI?5lp@V%6FdAM{b;Fm z(=ohyV^clU>=ywSTb);r)*Vq8+4-P4uRDM$OuahU}q<;r0KmR<2 za~Y%|Zj&Da5-;E@_+i;BpmVkYy+IghZ!YinWxzg?`h|Y$txu*V8OS^T2Bk7c2;>oo1uRY$tRnY z8+8kq=HMSzRs@wE+_o$&Z|8+l1H;t6|HxEW8tqepgHAt5xAaBKA)KYr7^(Msqde1- zY{ADg21;fVPFP7_regb9{wFFhVWo7Lit7iL%>K_*{8AUo4_`9-{e;zjrZ|4dlG(Hq zR&tlR|A}fzW`CT(v68yf#rFF*|F1N~P~492jYLW-?JHoX#J?&KWcMQY|0~WJ6RFJ; z>CF?F%@f(p6S>V3`OFiA%oD}T^CW0QsP=H=vSkIOCYgUa#9v0`0rP~Zp$61YUutLu zHME5qx=amyq=u5vK!s_b1~gD#8fXR$w1oz`OapzSfs)Wdg=wK616rsrEi{7`+CmFm zriDJzLP_YL!gNprI;bxlG=mP>LI+)@gFezhN$8=%^iTtOs4qPcSGt`$En!yZhVTLXpx?>b z+Dgdq`#GiT{4A&Rd4P~X;63fmuf(wJ{|L~3u>U&*^xuu;|Mv(!CWig*3edj}w8h2V zIQo_C@+WiamKuBG>{qtypQWu^&g|Q!@Ba@6=zskLwA-ft^8xyQbuJ~?1d>10USoJ2 z!tURU_KO;~E^m(SST6aH-9IumDvIds+|)a$C$!mNQ0ndv&V>J-3s>a3mrz4owVFHV z+6nvPWB;eG&*#F<6QXjHR;4Su`fAe7oltQ#)AgB+EcCyy_trsiyzjPea7$nySb`0K z;2zxF-QC^Yq0vAf!QEwWcXxM(;K3ad+&Sd?JNuk__r7-ay>esZv)D%W+xLaFed%GX^z#qf*Z?7`gb9r+JtDJp z?6B$zJ^U<$!j)=@gIU0Y%9RF|**a=iafKdwc2_JX#-ZsZb68`A9%)u;&-It{whiAk zyGNVRWygf=jN5hAjFxU5!prrkj;Bv5o$5mx?T2f_8+02ZszPfxJCF45+_%?szw5UO*smj(lUQJHx6LL^2ipQfnx8AxdFxpG!n0x1>k6oGr2=kS?nEt`V-;6 zA|9832?egvw`mMn^C43ZBO;SCw?it~PP1prq4K;7Rim_8=CtxiC}9Y8b_ER5)BD|_ zJf6GMpXJB@L}yR+Z;-4fo{e*y^{o0NdrydSbI$=7GqSa*z z+T&7zGJimm2>o4LvMNpZWA!*UI%FwveF+V9{v&6{y3fyinGV@%ktXDGDS}HMX|o6; zB5aYEwHjn~w4t_-)0Hxo$C2eh%*+Bmo^-JwX5rydK6MBp?`rxi)u*D;o5m>h%SaC( z=z-GnryW4rG!n<;H=noP=fil;hmLcxT+C)R$9(pB+?c(9-y7IHLlFoF`0 z4JwlVsg!-sy3)a9*`GnlSKD03-S5AkCkf(hdM$2W zpNPLmx#m$0c|{rw;e9ybu(e#nA<4#K@%mVMIrARTMM#G<>7X;AJ|_p%C$ zkS8NoZiX-7YjqAB&>uMH(I)Iq5?9n4$1K-c%(*i@F_z|BxxC+GOH4O8EtjXJAjc)+~6=t;r&CYnuQY9=l$XwSrfh_rftu<~W$*O@#hP{au;3*aA*~kw6;EAWZ>oYr7q}_Xk0$u+^CW{DJPK}&1w;y<52&vv| z$j?t@)2D2iwYmIA&6wNp|03<5s^8bH=DwH`CdpIA|tOXeM4TIfSn?O+SRMf*&A)$*(|)K-85mHX8x|o1;Eo z5q4I9Q}B>qEJTx9z>Ms7olvoW*tdBv3X>?`c8SrVh)4^29pjqjfSz58n({+|Xqxhw zYFb}?wtkmdc{sQkZwC|9WOCj;xevm^$@s)DQ5YuEdhlSfVbps zfp@_QCuXFU3MjX}>Y_FF)kJ*`@nv+@z_1T+znQNerEB7~RYI7Usmvfj{d3Lv*j*;@ zlFF%D?d5g}GqU7HMM^I&Tp%>w1@njq#_E-T{SUaB})!)`AS&Qvc&sBF3?QX&!y1`iJB6yyp^0G3L8EKg2LGY^FL@onv00V@dBmscTQz*176Ki zY3Ka4@qK6o+@KH>zYrh%9e>~#O?YO+Ie&2ihe2J%7~X11<%~;!GVM{N=CUtC{~JmU zp3?eaGCf9@no{lY*cMS=PZrOUYsV(4d8Hny(Euv0=pD_=8=Smn^XAF-Ofz9n=jV%_ zDJs%|HP`%E2|vJ=AA~KF`IJV$$|LnfnhgT58k(#JD6%Zd4f8H&RipsShSq$6T?yg7 zS*NN^{ka0pYYeChVkzxfEQ)5qq3QcY{Ffs+^ulY@%uRH+b!m>4;7g>y$HdHc!MXVO zgCC53NIQR>T(PW4;mw)JX!Xz`n!JnV&bm-=W7bO?OHR|z@DXoQ!n6`)X@mEXA1bwR2nZU`4rhzLwcEE%av8*3NNM4C{g|10~&8Yg-mSvoY{_ z(;6hFN(?REzFwUiQC*C+;V9n6W|kN_LaE=C*?vNu;glb0z{BXSzFS54_f_{Agfl?W z<<~3$%YNnhNDBZi>#l@<8LQ<>D-ZA%AWffZZZOw$P+?1BNq8v1lf_#$=sIQmIF9et zIa|fl#n5n*XUYqcMUK(a|B+bb+k|X5yG#Rt6=@h)<^^vpoEieRMEZ0l%Pb zM~+-RQ2UG9X3F~Z$kj^wrL9UPs>8VS2=V$TS9{YJ%vvB}CH9viBBv9yGwWf{3A|pr z$IbBha7WpDyebvJpT5J!uEde9DWT&KS*P*R3F7_H;EsHO@#F`maeT+|v|njb6L5B@ z0%mJ4=_enG+o}!b+ZxGFR#mP!(mC?mwxTk3{W5*yR2)!?tmrZ#Tm`_XTu_U4DZ;UK zHq#7y+RXs|`Ons^DlHNLlRh33`Ve+jF$u3_%#;EZUgj@SX?X>nWR7-{Eo0gqsxZsD zhwBV19bGPcX+qqo4+Xz>qQ-YgVs0Hn%`SR27V$S1TSLrRiryNhapGEO61J zdQC+_|63}u#rh%%_Y=H-YtQ$PE9tvbW=n}Sso>Ej;(FkxVQ8%bo znY1=SDZfuKbL2)9u;hh??1CV>&tR4iFbf)50qyG`yP}?-CIyDZ^H>=kdb|~l9c`h8 zND>9SFq+=DD?|Hxg)8Lym&i)0TiaQowdD7$PAzEbcu#@Rw(k#A@a}{|6=|z4&wbdA z-jNGGjq6b&?uW1mwiNB8;NivBh6ZaWTXtyxfHZmd1a~TxwY^VOZ z3O|z#SU9^RvOCGF&}*FnQaZP<8tBuvkP#^^(EXS~{FrVK;y7UAfV&OFGe<-Ur}ga? z{LOyvv&)v(A8gF9BRRehXQ3+pFE5SZFLLJ)hIQ45L1mw;J4J_>QTdftg3e_T7DhBMpSNi9bo?!4BnzT2+YhxOs{Z|#AKbQI zFdqU97=Z>6^{1CFF&L4?TRi4n51_LVbk;sbe`U^OyAWQ1 z74-{?plf;Ycc435kUJdIpOmPaeVQeT0J^m*bg9cdk}p0mX8~7+3D{kqCM07S>=;K< zwAE~>pf+?i(9Vz%&yWFkHGX$B(4w=C0_g87%2`7G^~vV_CRK9z`DTvx7@g_CK)8qM zsp;yo<8>O44r?~ibSIR=2e}h^8e?AaBBVRM_|*%~mC=V8^$y$b-5RXtfXyhP_-&XN zxi=iCckug-htm)oyJv<8`q(BSOKR{UHD~n#E)o@WhH!A>QJXi+f#) z;llz-64mEgja;0$=PDk@2jqZ8H$uC&nD@BS;BXMkpM^nIz7QkpN zd2RmJdmR=Q?5`Zri8cr>p=YJRQSjK$%$0u3mA?!8;UNCO>}i;SxayY%`d&n%bq zUi^9wuNRTqu^;~Q*v7y`>raUVge)0CW_jgs`jeM{LpFL8q8!(;)Xj}Zpqw$-#D&!^myP5vOYTas4jPp#d2UDbZD73}z3)Zi{^ZzNqP zBwdFznm%a9Cg|5pOA0jqttxw0!`adMsWX?jJ4LiGipPh&NQ}0SRo@GG_P~s z3OY~(u}*cY%FcG2fz z%PMMjAyQvm=a-S`Rw94xy*cu!#ZimEU$|;NbTcFV5ZCyZMaV1Z9%Frh))b5^cB0_8sA-1rqkzH`$W#xVW+nQ6^ z{Npk=H_pg`<;xe4f3hOkMcdy1#NYgbmfMuy z_nLiNGb;*Q&GSieanA4vXItX_4IPF8vWNcEFM#5J62;(`aCGh(Kkk~}d&zLy$#7j? zyreK;Q$+ZP&waX`Uh@wnOv@y32mzDOev^?g4x;?tn#bdXk5N>l#2U1e)gO!Gm#6Z{ z;%hj?LzPErvOz96lekU`pI$u=5^v`4G-peEU5v3}KuHNCeilgn76)k3Qi#$c>7BL9 zNkB|gVuBf3sUU{q{P`eKB2PJPp5XZbsaitRcmig+Op$zhqb6SQxl7K#wwarNpML#u z=kdI_yvCnNf>2W_(S2E~v-)Z`?=Ju2I%5)8gzklDqV@z(djdjG#pl<>%e>2fu!h^5 zbGs@4X2`rL^}DZpmw#w21Y{v-pCYUdoMOBwGGA1fvSFn``{vJHySX*7axQg$#v|!uN(1pW%JhP zNYWlxx-fiNT2Eo?V7P9tN4$NNrGbO?>e($YWv1EPMe(8HU90x$*v#J9Xr>4dRT@w; zaV-6Z`8%1}%UZqG(@mz!g&rhxrby-3c$SvUhQ1*);zw>;-!A ze7d?6MAM{Qi;o7LrIuglhrM;AXX>}~l4`hd;G-!!T8DQ_Y9|-k+8Pb@)2j6%^bLm2tKD)REju6;tcf^X(8_5t#F`y7Stq zHe76M84$rQd-!}iG0Qi(TcH0sv3j@%uSY7=Sc=&kgHIAVRe7W zH0CnZS5dl&)j#_GQ@11ikHX!5>3096+x_3Fa{r~<{l69l|MQ~Vt7i8fb-Vwohil4} zv@MG9D&EegYATVmEs1$3-tMPrs*&_im-x>uz<+MD|MRUxJk3`f{;|w9iW-&-0 z#}+aSA;(rSNT9?vG7O<8CF+V<3;bOTEAf-KbyT9CxHT3~@5xB$a(i~8w_NiTk6iP_ ze|~qitggZ1b~0_GP2&CG3JoU{@WuSMVTP~to>o!7B_F+X%dWlq)wk>8$CAX37f+Y1 zpR6qnHeT${1>3Aw=IQIp%{(7pm{^kg6I`0T)JQgUNv>-j&?G*1XX>a8mFn31ZzW&g zRr4j^_($_qQn;S$PRqJtz+AH$B>W1zLM$I@iB^}}c|N@GHeFlrv^QKeTDCpDNkKqZ zD4M$W+GKwiXXM6nZ84X9`=0zyk+$A$n=kyn-Iv%kc8$EV{2|MD?0m^l-fzhv^F_Tl ztCO}iuyxU$6Ugz;Ie43ZIF5xuERdyJe?2_BHl6fEY`!E2;i3hsm85i-&wIY28*)yj zr4cx|?G7$p^T>1`*^5h)q~9Z$bC%A(U>DM6*xpX^lq-5^!D z{OzZW1zk##ue3^cjQHW7(C{GDMwd!17GljA=)Wt=3NaLM{QN|-%14IQz^2hI)GE#vEUY=#9@`nh=-Z04O|JTCuCVo4N!f^Ek< zz^JLZM;#!}H0=={yl;mrc(9D7I&ZN`$zqhoyvEDwgxb}x1z+Oo&zpz*tIY1xHX^fT zJ9$VJDrF>5(=Hs|r9x)?fWiJ9R%dXKdXlsuu1DK1TYo+zHidT5YEf-fd59dJ<+ z>-Oy5YVTfUbtKqm zjS(C5PG`+0!LEA_W{I-=1L)5O7iC@?#|U{C-XW5=pATwa4%yy+H+OLAP>=h1EZ*-z z^RBDU?PfA6aW0oBt>Vn*gMHDes+c8A7cPo%^M$I^8s^=Awdoh1wL^;@nP!TbNU}3s zrVi>FNrQVviKh0pIRHx3(hR?79(88=-qk@aFU3>7X&9NV`-sT>*T>ei<^by75p!7D zt_w~d1CBONL7^(Hsk(7RO(&~XWGop^ArEMuzQmkK_o^j9mzp$@(CS?{KKY$yfPW;s zQm4<9=O^IC= zN8fY3R035$YH{+Yb1f`aPV{))Zea%BW7G^! zFlF!cn8TH3P`@Zyz|pixsjRI|hxK0o((5d?KP?zBRF_Y}POhzCNxsV;PR%X)D4961 z>^PVmPz~*RLIX@k9|&o@nWsYL14g`EX0(=SKQnq+f#)X+tX-7)h*TOkCyY6W-=U& zks5W@=(CXmzZLjCjnO)#%XdOIoL~;u)!C$fg)22*rZ%_2p}N)^ zS*ZxRu0)7q4dSXb$bxxBhbs-nbaN;8)BnooHbJ$n3r)*v`4mKq%5sRWU%)S*?}@4wib zH9pLMW8?gO>(#D)Ail{&iQfl{J>*gz&1D@k9aAn?Uc=jk+$8L;-LL^$vWXhDX}W;J zX5s2n)2t@Z{bgWBR=tjODf@E|t#XajG>Z)|ry1_&{%52y+=&My=+GU`ll6E5Zsra)r ziq3x1Hvw8(Y4QRc^0IQ8i+IY#vW}=D+-)kYaO!mhsTu?lFXqg zHMT$?Ja!7fY_iV!X@`j6(b(ylI&tw>29H@_%o5I(Sqpu9iQ#R0|KJz4_y7}2qma;7 zBG$rFj*?_u%f}NK3anBAsranCxI>u~Oq2F~kxz6u@YtE@Z{bSnF}^bjA*rdkUC+_j zTs!=@A3a;L7aO2-@mA{54J=!6^zqRtE2;-Fsiqe_8KtmiTKMG}igjAOc_*;%#r6k= zRI2rfHE%d;dsr!l9Loh?k8li44t&}2zJIXUgNvD6>;&B=N$KtO2#OL>zEA!&RhiEE zZupYTyLQ8{)3+TLO6}sSm&n)+>5cpGcJgpm^QUwg4{sFvz^%jA*=l-EnRb9~#V(+s z8MjUrd}RIYRo8R*UCfHk%nEJO>ml0@ZW$gg)X>oVfM@$cI`fHsB+E5nu@$+onMHfq zSgfE;vvj(9(qcEaYuVT#9bWHH?>wH8)gyKy&4LB7f(h`97x0e%@k{LktoovPK)gUu z{1wXFDXqqFQ}t$Ufm3Gn;-P7SVjobh!~qYbQ6C<)nlfqwt|3hP4q0|pEA~PB(=~X^ zJ8CWGz4Ad-dW1>SlUpXLdB(q*UWSqF>FCfT!}O*+Hv3(=1x~Q5)sKQ5qs(dfRncPX7R6g34B}N5gt39xX{adC&%Koaf$} z7nH!?< zcjUz7&(pX?LlTu{^1M$IcueC|A>A@XMM;u-rvVS8IZtkm;zNDDjM9S)Y1u;KQb?)R2<86n3fJ>Jy#(U8qKtsJS>Tw`Erz6_y#ELM>}yLRli-LHg^7D5yy zw{BZaxZu0hu_Z~SGZ{P6CaidsDBq0yA`WD8O=}@rvvfZSTD-1lCdAj>NJ$V#OF-u? z+xqFxO?w1Y$EBWpGN%O$h(nz>SmXF{sfS+O_CYj{WmX7fXu%>Q*IniamK!zq$c&o# ze&)Aj9fC0lMJN2-!f;nqR*vaQ9>e;?sv$Tq@2a|j^RBGs7_D^-2XVMOII6%lQEPN| z7*gwW+VL}%EBz@tAJ|a zNq`GmgO}*`XMNC|N#yt;6XH>|kNN7tQG7d91@>2Ox3w>(QA;KS^;#N;qRHT$d>HrF zUc}Bk4O7E5CbUPTa67X!iM8x&px)&{@yg-A+6Zq?uQ{iuCBqcX5Xz_>lkAS4{7H$y zB)#zGt=7zwBVb@^RiAq;zMWDC`L@sPoSIWqhWEE zBH>!imrMO)-Z-fx!l_(Q;Nh8P=a?T@_)C<}qU*J_*F5kk6#C06d`Jh64w^Os@KP;%EtTC0M07wkZQ_C5(X1IO46zn#K(z^|#^gDT zHZY48PiBcJbgBBA9W0cCYJIpF&oS@YSBXAXsSIDS%$umVE-`FwC(J&y<{Z`R#7y7C zE=@XTUZM1a%CgRxWwhLL8ud4of`nA0@_{sbYMsp_KNPViYc@%w#o^np_Tp8JKv=v` zSUeD!rOp9Szcc@5^k8bZ6UDsq5SvfP7clC6y{}CfqD{FTLhWOg@X|gfv~7YF(dIG+ z?j+V8){sN2jqglCvgGnFAXdcmY*HT!h)HIdoidTHMT$(gbJU^#AYc11yREIOGLqeu z_Z;vL8X`KU!+7 zs#o1`@GA%>M40OAebMqj)X95i1>>C%GZ~yQbmdpDjpdjT#JrC~^QKE-eU)^ibnqo@ zjE`LhCM~B}pgYL(IqBmHFfmUSuM9TfE| zsFZ}MdNZzc%c*)3Qo6$C+`mfxpY;H;#Zz30@nz>Ec@V5$8+p!+w3C zH)Pkk(x2bO=IVfQSM%*F#@<>n{j7Lbfa1Z>CQ(2Xu&s&WPdqL&8%--6b8T86JHcD6 zMg7*JFIYt?t~}t)kQ{Ho<8~DAxc$6M9urE31B z_7<~mODtCK&pskW2qMKg0+f%QPTU3E7eN`D73emq#t8=@c%r!`H9}YN&RcbWM3ZX4#E}JqAJn&Pboo&~sFD zmqg0J(DSM96d~U!Ze*H=iLZv-6f%ue$-&n{Lok*y|)X&QZr@u|PBfrpU_{{Yt453?_2Y)ou! zf-=)G7q*)FBdQf<*D>`#!LE;*jUOh(O#wd#g#Qi^IZg;UE;9e`pvg{C1yw_aSD)pZ zaIw)P)Oqsb)SWcyopd`|=HkAP*r%VdN46x`hAG@D2_fT(zpTPALr@pI7YM>%qSmH3 z5=w5f)6>!}NkS}paz8xy6*xxVfJ=wOjCFUS=C_}l)Hs4z?f`T3SRwV;koJ@}A0$!V z7kgeN<3C5rN;!&4o%@&%c`858eut4ueCHuSXoJLWHNwX0*Gn4dI)HHU-GEx4f4*Ay zz1S>Qt_w&(RUxR3Z3RkWRD0*mdCW>?;&xQxHi^3gs)qz`%0GklR??7G(se%a;(xV813MhQrK@NvT-Q3!pj)EUQW*`XcOpQ23YgbS3l`F&3 zlE0h`7fT{wV{C=$VY?U}6uJHEP8#A)8Yxx?wM}wo&Ko1d8#B;u`G!NRo%&|~dVafb zI~RXC!W-eloms}un`Ax2&O39jBDQZ8kPBaVvS$ZEvPjPO#cYTf$e@jSwqQ9!MoGwP zez^YWtQ`+*6=3;X(rKz##H5rYw3|nZv0$S1+;U*y(t8P6?XBGT{7^8t3cwa2k+_Sa zdWfG+2ssW4olkL-hH#UvkMbNwzqylUx7Fp3ixG;O-r9?ioZ2G|!O9>MXM^1)?c+_` z&a?)g&l~Q@3#w?qWR#rJ)s~#Utx{>w7Duiu!pmw0D0uTl7Js3%ZDBl>VyV&>TfRH$ zD^B#%Vbh+cq>{!eG%-*kqkC9RAHO0dCZ!yIR1xHtR7~?I3G$1A;J%YUMwAHiEuMHd znovZ#TOk$XZ&c)OoZf1P!6Ka?)cFLKqPzwSC|@35qwJ z5JP2l;Awd#s<=+ zBKRQAKDt6Dh&aEm61SkBEYvp{Kma`z8w5t4jHiEgK6OS1*HLm>*8I~X8xu*`cAV)8 zNfnaOI|UOl7iyUM5Hn4#`8J=}`qSgQ=dkDQ-ReRf-r_yc8P33nhv`DU?R?O)0}3o# zFfMunzoXK3GB!9;Hj&Jg)?u6yriTZfBL5E+MxaXh#L;h!`4FTuFIC1n{R<$XNfc^{KDnW)X8Gd_kw-g zi46bF8R#Nb9!{gx;xDfb`+olk5b+GBz^-tTy+CBaskGiIygg@ zr@bn2;5SsWp~>$^`Ge0h7x~{P%fBgH7f^+gke`}BNb!~q-@b3hh-=twtuo5sfof97 zHdiSkgoN9e=X#YFvNhBWZgEBr*WfYN&2vi)SRE{ZPrQb z3jHukv(CS|LNj;FI7BRZ%%3YwbblS5Hh%JC)m}dd)=>Es2b=ojhD8$Ei44H1^24fr z=huf5)`tu3^bA8cHV}ynUhRllNwy;9fEv<3Ctwr**t7(5r&X2O;WJr?6wnv|?T}6J zlx64shYX;M(NaRGe(golSTr*iAYZX;kyf7)%SVyy{iLP2h z{*Z{P4at#*1aUYoBj=#e2Ad`o6)u=k+amH&$0VH7Q*!qz;6fQCR%4~K{+EskQg!1e zLUa-VT(#RhDy*nJEFWyQZB0;+V;|U}sEGw#3=(t;gG9+i4!;<{w0Kf-T>%t+4*hb*o!Ck*CK68 zo6a6!IU3H4Mq9q$Z2ZIO0~nd9G-M`GD$>79Z;h28DhXGkbS!!7b%kZ&CWdZ%u^F9r zPJTcXHa5w>TqtrVDB>t6Vc6Gi^TDh%@INcH6ha3TXuz~IWCQOCxy!LA8C#rcc#R)P ziM?qkZ*O_a?lKlX>n4tn5`Uwa`0H}TcK?_K5ygRX9h8*2V-)gpOnP~X*6sNhXzvT zB@jtRYRDmi7=~kxZec$WREP)Jq{8gl<$~9qy`MMb_D%(%Jj{mb|7CCmc_k2BKfW4V zr4354uNbSMrlK^nYJFhn<_?`pZ^Z;#H6p|bsEOBUo|Fb8L)k`p@*?7aD5r$ujW~qY zr{?**oSP9Xw6L1zD^o`p(ENDNA2SyrF{N5o-;ww<(t z7@;8;p>gjtis3j~J;4Y(AfiH&Ofn72Lk{ORN1JX2ig?8d4=d919(ep-CX}jGzMtmY*RRc|fx!Qns0dz{HOR z*A@K)5Rkz0xF|yh&Y|gZ2#me%vv}}+(0-@r>G}$wo%DlKoS6RR`Fxk#?!0(t@|v+f zXL3TJ0AJB?9X7FozE|>x$%yeZSh0?Ho5aYHI<-{g{U|>cF3NPQ)?%iU3n(2{a4_S| zTm8x-e-9jx(SH3j7x%_y=Afk@HwNUF@cgDW=uPdbS0CoitSKCDox?-qAK`tY^X^wt(Uldx&Uh|F!l#SL&Bs z4n5CEae?nb~Oiyxo-de&Q*g7VX+rg=ovOE-6nI zP|pnlHyrbyj>}JKB6G_`H)BsXfBl!U^^jVKps+;DNF%0O^}*N@#g;=5DbC zXeS~2)!7=ObsxKe5wj8lZjxGuvkP?`;T>*MUq5v-FFWnyF;gh7o;Vo&LlYK++3+@! z>(6Cl7R2>r-19MO2~NT8Ajq@B0+97@t6OQm*ib|Auv`4IRu-LAz3JPpDtM@+INP8z zX8AosRSaX4#vvi&FnO@8K#Dc8O#Pj8Y6;LY`v|{>OT@Ir)Ax6Cmx)+li1KK`Wz;N=^8m&Q^k(C1Q43F8k&$ zpj%CFovEN4P@uuiAY*8McI4!X_ zEqTZcHlpIlx4-UIP3ZsbZatrhrJ^2Wp(ZcCv+-@BQ_DXrUf@j#u&+7%wLAWG-84DC zujbPmvs^?Gv#E&3F;0|iNk=_VQ_loPH50{m0Jeu}im_n0#ytL8rx0#^z| zHfo?zL!eOu-M;25COp8o=HuiTYN~Ob1XA5=Yl!J816}R7lEprnCwGHb{8LS(=aEGx zSBx~Cys=3bhxD$mmL~G~Oz~F1kH7DOCZ~L`Kp8IM;*z>eQ@M1*gQ*S9q4~gd zMKQa9nj_U`X+HYBp-&dYig2F3SJ_>Nhp8ttUZ65nl1WRrDa)NmRq9YxYS+q8ae-RR ztEt0Ro8)9PA&ApR{t$rE=~0L^wdVbACEyx+bqw~Xv{B@ zzh*^K)gDW}u_3K9K*<)_al;Z8ryPt^{@7#>*JLkZzp`LAY9JKQ76{e86WT@=+?FJiPXh~= z--A4qH0n{KuB9AiAr{=u*Q9J4fvruP6n4J9u$Tr@N>)qn0*48M#^-z|@2E85U1EM- zT$Sx$^BsEV3lX#*Oa!Er6GA3b=uzCQ826?Bc6=QENxa|TtjK;ciZ%k9^aHZEN^t3Kg96lp!$3|_iXHrZ0YaU-i+$wteM{KLa&TlhY9gWb>)e&crAmFu zI)X83M(bLcn=vyM>3IB-Th6`%lDNW>Kp2sX4;Z8TTBzIW@FGNwug$UCE+33g&WXBG zi|@`V6DBd><5qHhQ!O`V`S^;-%Y@ArB3A_%pKr@~5Vg(l-@F$#JMjD2@G-heOW;)N z&S@RlasAN#Ln~Hqy9(NaE8auYgzB6aUmw(p);Po^B zTzANSI4=0hCIHh9HOH1RNRF*20|`qj{?V$HiKzfkb#iPH~kuvOW8>6Ph| z@Wp!gvyq`LGszHB)zD)%H}05Kp4<{FoPRsU@taF?;@^dEj14+j7=_I_C1G2vSRs!R38vp zcm4`&0UMMVej2sM;{%ndw(Hjg=+}i2sZoPD%3Nb-(6X^WNvaP}GdyK)tiXKxDF1W= zm|+P7)g6K@YoRf|Ge}UO

|UaC@d+1m40D!5uAluK^j4jb+hYp2;vFGFXwV|W>>1;|pgb3(TAu3%1hl!hv50F@#n$0H zKl|Gr#w)FtJc!1{y}sf$$ywyC?neyWX{562Rp2zl9zn^6StbDU6{b9Ux5>#S{$$fW zs9L{0tV(>4DB`-)yB`e7Lgn)FEJ<Y&lAhdT1k` z$`d3v@oRahkS(RxAjzLZvMIP&8Sx+>98}LdBoNfr$q_`|-~t zAtFDU>?oxzQ2PVJE6Le)a1+NSx3Hm{~X|H$!c& z@0!@?W8}iB+rwzn_z?IQ|Jq=0?8*wQ>AbNQ!gQja97!es(nO^+{~+<#*K42MBF=Bi z5hp}>xEa%*6hbNxz!7)zRj3W)+W=}1xB!0`-5&%SF9gVQ1Z6q8^Z%<-5OIP6#2wsW zAwT&*0dE~{opCooOufW?ASJ3b8#%i=aLDHy#zx<=<*2{=0oG0LFh^o~QN~|1=x7x$VKLoA>V5?*ro)RAN?q#h(nH z*>4C`>=)q0F@O*PiPixRI5d@jEC*bS;#t0$NmnSFpF~QfU<3Sn@&r_3eSBBHsupO$ zZgy^c(*w|_{%m?fe8FD!W0L-mWfe4H8+=}98+^efc8Ix=Ao*#a`w9Z}c0`W=^q!YE zAzHj2vm>reNiT^MOi+Z-OYIo3afA@C%N{i9fJ-3%0RTJWUuSQf%zdD;_ozRcxf=Ci zJ8rVFhHctq;Ea$K(B)X^q9&5b2WR6e1~CyU$()YiW~b~%M%Qehy1BO#1&$l zcwnr9!x+d0WcNVO0~?`fKg397Pw|Hx0x-_`K<@W5m5t zLdZg70Bl#a()uCs7Gt!3sl(5zeic+Yt4<+IuqOWX8cBXncKqmcj-Xrn{mj>LX)iw6vqlB=XQMI_&CBFhal7%^%r7oh zJnCeNoG{6V57GSi94mNYw!X<5YCd$Vl!}o+Cz)83Oauk%hoUJzao73ph=xN`vNVf2 z>7NkE#mtt8rScdvEk1w1-zPY{Vk@kKmd2#bOEA_lpqWOF~iPCK)z zrd%)n{wlZHBa`(BkwFX%nL!MsA3+GE9|7*fS9o)w895Q2pVCtqK9SmHSr87ju~P^H zCO+YjVHM}bzY2M?K~gfO)20h8l6H2`Z`@)bv)>2D@CA<94WDQ75giB6y9m7ch0uE= zNOn1_!oXPgS4F=VFC}3Q*31!By=wm;F|LJnySx3EI7^-}@!PhYjc(zUi7|@FWxV4l z=|Ol8Pnwz*=0mZ%35B>OO47NE7|;Nq6ETVtd*Bh!+1SH&Er&j!Dk4sbKFSATc2`W1 zke>Yj?|f8?socZ1i415@eKB7Sr`-Z{I(!=kBr#(A8Oh!azVu zo^OTg&>yNP_p&j3>srB_FCPV*MWgn=Z--7HM~{mW+Wwu3Y0t%qNsi{8H4j!csSPaV zUNR59H5T58C`uBODGK`f zeRkt3l`%2N4wo|V3~rIoI~8D%L^Nbbtyr&R)Xc3X@hY5bwZNmw_rz{2RqwUmZ-q@d zguWf=dD_P$nZ{J&j(a3?xolYFe_G=2qjllTG8dB#H<%+Pp4)61?$7v+c00uR3Y1#D zmjc&+9q~E4v^={q0-?ek*OAv3cb;|H6?CJPIlgVro}PdLWXH4pKcFkEtA57G?j+tv z-)GUE{g!;b3a_vx$xof%<+J|smttwNM%EXi@m$pIE6mnV`dlP^68hp2ubtQdKLQcI zY`7HB-04zLEuf7ubVY0N>X6s}$02`Vm>JROETrf1kZhsxx<4C%w=Fj~ zj@WuBGKv(}BY4txqSO0sW)nCxnzx?v3-*`iW zGlg;{5(__DqXOlVTMA5f4;5upcCaw4*yf+!TiGh*%N#q#k|_`R??U}W-v7|F{9BPk-jR_}U$@9s|iK-k2lJSqUZR4i)Hl zi!WJ1L1#5bE*lp@+ik4IM!6CwXO@DiR)Q*DD%R&;F?YFHYAElm+;Nf5Sd3v_*vQ~7 zEclume7I2DD+1Xf=eNAHB1`5+(6T&07fQKTH@(vOd$<|zT2}j2d#z0>XngIKZEhMY zlE<9!wc3;S9ww>r{P7R@c~b34Vat*`{~K%X92{BHhWp0WOzdQmiEU$I+qToOlMW`B z*tTukwr$(S&G+4N&pA~Wf8DBH=y$KR*Xpj?tGcRx@AHfyv36Yzx+kIL_QC4w1ms?h zu&pLZs~tW>9o+fY%WMEQyxmhGw$UF}0uxt4ycObP%Rq5!m@OU0HM>PABBkfM$4wfW zDeAh4o_46-7HigEUSpiqAG_h-H|q1;<5U5Pmw?>PzlVVdk@??nfd%I zNr7P#mc3fW7ZI{lA+filHf38TQ2XQn)epY$R6fI-5jK%vz(QqA(CRj0sdZ9R~9AX=?8NVmaBPx98Bdr zeTb9)IZVx&u<4_hMCp;0SeE2LmZW^urcE6}=e)0I^qaxdpV5UP=R~~l$px9E?H8KK zBDM`O?#^tK;H!ITAv1o~a*C2i0oc_p7z+5wu;(Cc5iv6)l@BKoZ^t4KC z#VY}$D-(;L;XV7yilx9*N$tg1CCM`(t=2)sdGr_Js0kOVEB~`i28MZ`b7QNM z9e9GzN8U8%UW%(jcjW4)qqyk9wxO?6Bzmrz0vciMex-MIb8Xep?3GLIZQ^2MYHN!B zy0o)b^kxAQ@ɷ~VW`;Oo+GVLe^h_y2X+{lA@c!2c&HzyF^OyOvZayfnOqLnMhv z)en9o&eoMjgOQnu#i&1#wm6(noUImnFu?ue!hd-&d1*mNGLg28Ky?SFx`!{){bSK| z|L6KWdjnj-AE&HdQE4mV zg0M}bATOI~<#7L}RU+=3?)-LNm4@y3w^PRV!-p3{@4Vg1hJhDtVn%0eGy~S~mfunx zA%by_+upzHi*hoEgKT+6WN1$j;$+kT%lfQL5?dsdY)dmu>~!(`y;LdD>9G0t%z_^) z?xElB{V7gcQ8629vO`y$fW$0pi$)4@c`owoi>@&dMh*tJV8A|IgBVN&(tG_YN`8~08V_qFT zXCpbp9IN1?UN*MNXal+L4Zx!`eCmI1;F07sDyjY4FM|sjC@Hot%kH8SgBWEor_}d{ z{V_mc=+ny1;AUlI#rU*8P=+7`zRZord*#gYYl&W#t8nyv2PQHDf~d)0>6CG~c8mP5 z$D?tyXlG!c4*2!4B;ji1SqOY;+uj%@GI4QOOoY@-qI@{XifG(8I5Q(eREwrleLXtX z&HFpAGzTmw+}6%PzOx%zqGbK7VASWV)+P_?9J^1~Rr< z@KtuDl9`w`?Jlx88|glFnld$Gw01bPKtmQCRTL_%Ix#CWc{(1vs_JdOCaN&6p*q6) zcfGHAGIYK^0Q5G?RmX1As=6`{=5F1Oc4Gglm<{Y&x#DH#SiNd>>8!=?dT$DHx{D*u zuRtc90;QoXi_Wjr@LsPS%f|^1118)TFO{=_*(0!yCAl7-Cr2yVi?w?7l|O2-dZ|TL zNAwv-@OSl=kX4O?+MREjTfOaeY%gpbvY{z&OSccsr|%U{FTqYXOw`T@X(_AYbX)J~ zG7qllGKK$Fk+gU z@Vp*>+B9%f1^_1YlIafM`|{c)YIl`X3(4G9D{Gp|dY-O)UN1ptF-97YUUX?jgWEgeh*tM@8l;G+eXMKA@@=0_NOb6{5hR$TkbO`WKN!*ymmA zYiRCCI&hx-X;vP_n!4;R4TQITd-u(GR^C27=@vSEl&OyJ>e|1B@mow8_}2Z`3*{<| zpt0+VtxG%UDPsPjs-vvW+tC~N+4uz!nSt*J72LbC-37U3aoEIfeP}C%pHOxAnWppk zT$(1BofySs5q6nU|6yDNW4b8hycUduTH&%^Wp$Za~J1IuB*0DQ!JDQ3?%ZB&j-w`&OB;zZ@s7^x=Eh^lF&WfRYZ8&WfoTUUZn@&!ubby&XnTQOn2YM$(?a{*ZxfQywNoJ zPRRY%lo}g|^594)UY!hq- z^N78qGh{H-M;H7fdK2v=;Swzqk^WZ2Bjt#2wCW+iwPv)eq2p6X;Uji`D`2oX17fYt!PH}|97LIAy>;dY zc>bS%Yw#9>Q!-q7tHc@Q57`VwpcB8RE3zP7G0FiG%G2+8TI(DoYPn(exTBXA*GWyL z%_~QTUmIQ7ZHZi_P=-@U76!JL7T%5xHA;7N{H7=5o$it=lo~wnS!`QKzJgvj(L3uF5W_SHiP{j86EXi-)cHSuHebSp@wQmApOr ziHzz@!z95r6yvO@Q-zX2d(W6m!iBE>r{^SoTEy|0HnT5wg$+!#K}u7u?Ae$`?4}FI z>dNaCFCc-}tY2}9^J~obtp;1e5QJ@n63v5=MugnQR;fd{PRLxIB!3wO_TrXQp(bSb z&kYOY*Ajeim;A#;!NM?HW;GVM89EJ*HJV>I+=onQ2an~INEXP{?Qat?(Pm>9geSv@B(wopdF(mOMhVeQa>f zy-#;@8@=ssyxgA_{@LS{{|%-<;EUOVUc6!i+Wkas!Tei~Hn>G^m$e{RTKb>4Y;0OG7~oHu&a9x*2F zRYngFf?8Ywx}QK<;CSIluC$yU>zb(Mvflg7L{!o4i^laIy{b#KWY*{UWYZCCXkgJJ zjx+m<-uXMgj{AO0PZ>E@gU9j_zcbn7;F~bep^4Rc!?PA4e8|XHg1_Hn}z4CV}hr9 zoS?7AV39K@1iF1_ugN4#6&WW5E70r8-UT)`%vR8Sdk8dCMA2MTR@Aj5TRGEv=%7D9 zw^sTPXV%pnNqZ@5AkJCXhWkts=zDqFB&-%OAd>7mA0{AX=O+Y{UHH{zNL$ybe$HnjHK|r#8gM#oS{SfBG z$@&Uv0LRH93TXgGhJytG{~z~hYF@L)f(LHhUZqrx3y0FnG2_!HV~$_$m*!q7FPVm} zu{D8giDvicvNBDk<`H~|Q!jQcDTb9l{7Yc)kpI1@nce@n*HF4*Fm*`VebTeYC+nS7 zpcQD5N+)||G{ID!r!@-vyj|gy4zDX@wVYa4?V#W*F=K#J3mK2D({k+b9}5i#tOe@M z`aFpA{<8b)a<>%mFl?iIG1#>_rRC@?=lyexB2EWX3Of^XVqNSJA}W(BwV})Ms;r~KI~x4Wc^7xES`I!jn-HI&9$Qb&>kC(+2$6BXFq z|4lYQKw5T_%=8-Z@de7g3ycRhM2C+5CY>N4DZ5E(dX41x;>Z0XjE5&ghqwPGpCBMB zyGibZ84^X7_^%Sm4ymYkbTkFA!JnPKN&-7*iX4#WvIc=l`a5Wv96(4+*_}Wo-yJkV z4oGZSgCM2s9W+Z0NL<;SAf@gdGzShyd|88FrR^Ow4-QB|*_~h|kXGGiSz$<<$nzV{_hk1SNQ)U>VV9Cg?`$j zibygx!f`8d8fgO_!g_a8C?qqe+np!~H79rD&hs|SmPNFTRO8a^NANf}fPur;7(;hh zuMMBr6=So z?Mdu^?GWcC!>Xz-?IB$|h|C$RpMc%LJWyQx>GW;IfmDAkMmEC%`lsj+q2C*P#ku0v zX(&PrrSpd{@gvWP8EWEEB}GB1LChh|;P>Ai2`$U__6C?;tB&06swUpRGfuIyD}eI) zoCRetYol<{RYJS0D4C4vnTr9JMyx261V%{4?07qrH2D?GL59RWHmFxs+ChCPytDY)7)Z8(`V;I1+b&rIxN!Vp07e+?+-4 z?+_N9QCYl{;-ugEv9E9O^4wLDA&IB37r`mqW*!%z^RhO#j_zQTr#}Xg_Y`4( z#51>>PxF_-Bhz>_4sZc3b!c9`f!&z`F1=1po-a>xfjhPIwUo$v23!J{HVvA#J(bt? z!z3NlDi#Eo{p+^8gEpOkN?F8?i<(}&>{P!5ptu#X)u@9p@{)6{@fT2Lh|()kDOw@5 zY+7re2qtR@?e7wh$4WfwHpMVuxqPQNz6xzj*eI&;e#%CQ5GEpxkHz-&Ng92qtoW)B zIWY~5#y%T)Vkl?3-W690q6)sjgqfmfK&GuX3lA3ajd;#3V(QOA`I_1AE z$_C}VzHKu!S3BZ7efJ{md#9EP>KA|CY77HVP=YU?waJ+ zF_ghG&2{y%=w795#(G{15{4QZq;Ql8D0mu8;G-NeM{~P;N%R2NZfi`GOKi6sX&rWK z4YjJ(m1GSOh%({H5e`kBaofZtMp~S*q$9DPXj=3_M|7#AK(p?;X@Hpwj!g2*zyyoCI3= ztdTI|rJ7ULs+~@%S--j^e|^6&In~~o@eY^XA`JC;D_p$_4*W61>PlpDWBv)$EjSt6 zTlX2MbST*QjJcN5+Y9{#fcGpR)11psc>A=Cy*slplmy>)ZAvofCd5uKG{%bEsz%?G z{ykrhyxt)Ik0w~s;yyI(zhowv1r8~q8%3qQPt6zuCXlK^U5E&t5n}S29D>yfxUV)d zx7cIvF|MvFT4G2MJ~)3#0Dd4BmAVm-j0}n2GRAs^lYU{v>YGH~)OQEzj!23$bvFZ*OXHbzVdynQnDn&_z%`n>&p5bUZkxQ&n48135Yhn`34? zDHd)q1aI&a8)&Y2`qRJ?liHieLa*c{gN*RE%qSGs+J7t7t+G{EKjMLA?nJQvgSQQ3 zf3KLtEAdmtUpSiS4pt-r=pb%AP|43$>gIICZnytt^I9qr7#ZOW`h`BMWuuf_Y~)Rn!b8{rnLJl22r}CH47>W`e|T>mw&NQcIFmX4r!}-sB8ACNm!p2&z6Qf0c>5R1xSIeG<;_}> zPC1RjLKE1{-S1vrh3e)XUt{fElI_&oQkT5M;Ns;F`UlUn4bz0z&IsC${w?Ml2_ zOgS1lEPyiH&1(oauSY`VK|%VuWlZ8FZk=lO<1=X$2AK-CP~-2{1Ka+a%{5qM8%yFw zJUV^WRMk6m4g;fudxM>i;J@^sr_*l&OSe$`#1-sG=;YhzMb|p^b@?-1qrZyO{))JL zIUntYKkZx9GW!W@vvIh{=xQ>}0lO?pjtbqG>dQRyEBg?*;u%g~YUCBZoNCnL=|x|< z=jl1Wh61zc=(*#4Hi&?t@AhW*;3omWles)~Q~b1~aaEID90{q!t9&866KFl5vgFA@ zQQFr-jdeU5Ij8K`0sXhvQ-C9{yBor>&c<}2!z{osrLIR3>Z*=qo+=ZtHo7QX_itvGEEM8QT{?mQ zOni9KNUxJXzrXUJ>8W^F1N*GMJfd-mS!5Rk+i4ta)X zv-8sdb+tkkpvfmtoW;*qE`1hJv-v-2>aGd6!+162ZXONVmvk##sA$^BrwpTJXFL&w zsLX-R&t9y-S=TfwB&Q9&(X2))vF`h!IbDkI&S zG9#z{rM5==G8#UfcD>h9aWaaj$IeP$52vdM$r6>tR+~NMKz&BiCJSf zx((@@yDGaRB}Q7_ronw@Q5Pa-b9tj~OaVZZE`a(*#GiJ197rK#McsUvQQDd3j&QUP zVwy|$rmB~kCWk(lET)YPq@e}to7X49!SR07{q^@S&wYFklU`f&P3@HCN-~i@OvYqU zBIg?uG1$HFXH9q{yzx*(Y5@+u!l9Ja*^k_JF1*Ic&1n}I9P6~pvPERaCosb7-DyCk z9KQr?59&1fb9BEjEtGJ#S0vcsn zST>6O11h}-A6b{D#|Uo^wadCR7lkKC&7;)8puftn9RV#t>V?{SRjA@_a$5s@7TI;h zF&pn+kCiqpHl!n+Piph#hN`h@`7prfR?XR3*-?&u*TJIUh;jh^Z<1C2V!~&o`4LW*ezy<;4SIRaI4Dxr@s-@4@6FLWIB7 zgD>Y2;ditD3NM*h%{^~?a!mW>tK$D@pv5>JT36OCOPX}ewux{jCXlSy4A1~IsOQxx zibv7=;$TMAn6|KMbBF!gqm8#%V_m|?T#qNyJ&~;mQbB(Dnqs!%k)t!AIv1)k95!$) z)H64*tjyml<*eubHh079yu*@uai>nlnE+o9fq~mxS2Y#0nOh?`EqCAhW~9G!2Dm`6EM$SlhrZ-&m=FJ$?U^0Sxt{fV7v{Dmr0~pTq_|zX-LtFJ7+-#09FG zfEGdzH>tT=f$Id;)muaBEiU6`R(R0jH7wRuNqTr!Y_V$C5)JS$bwc}$c?oL33WKE2 z&{#bjAuPs#k;z&#Xkxe;8=H-!&(!!H8{1n{G%IGt>_1&_(LvuBMtx%htp9X>knDxT z*!K@}7#ia-GDTuz4~2&ti4HQ7>?Oz8_YQLy7^gEZ$zoxXhleYP4pNfr#mCr>4s)0o zr!z85VqpW#!o!(G2eFF+(sQ-a3$@ZqwbCoK%4_RZCri1@uF88IaRByI0DCTgy(GY1 z6JT!%u=fDihXU+V0rsT;`&NMcD8POdV1Ej*e+Ae>b??&n4y1sD)q!G|LC{RXsi$LC zb8ze12Yp}g8%O~QtAoTaL!g<&QBUWr=J3?Vw}m;2<|&s}~=rx8QrKn$+p&00*cR*=dQcgbP%6yLOybl!KbogH-?A zp{ZvPAy%v)SAGi;nxMIYYkV}+zLu0nc-T~{b`zn?Z|Sk_up$X!N>5k(P}?(H$TMag z>DOas0NSK>mr?|*F^@#oYh0G*LfH0^21KurV~9v-x2L`xmmivov;pWsbD^DcVKpUv zmshdp#~DldA|9NYOQwe2m77cCdf(^2>xS|G+{?7s#aYf{b!;alY{Lu`qx*WJr%eRj z3YbpzKT%KvPpCOX=q0Y$ycc{7mc<~@MBI?T=$|AG4*hJM#*%0Ak?|hPC=I!Jn=ryS z!t=O35h`doQXD#>ayQHoi-w%z_e$gPJBBhRg8~*qm2KGanl)#(5NL?DI!f8~yE7Hi ziYIDZ6(SBq*v8Crss0f+IU=f7A;`3ynj~K4@{6gvdFX;R*n|c2Xj)?j0uJ&15K@>x zt1$MUPz>i{V&!@m;fUGjbh2uqs zXY_MfL}aNaN0=Pow*E9Ej)ESNLWWamS?_u7*Aw8-Bn~pn=-06P{Rb6RKq;+hud0f4 zUUD_B1;3BtkoqgXMcv5S{2B+xNGMo|JB=l|u7ng_tOxWx9prlGQt|M+)mlecojX@)_7Ya4E#lM$LeDS{uF6$#DYU7X$7gEWKXU|o&zasJ&%MtnfEZwtu&9!$%NS9Ts?7KV=#J_qD8yD4OP(aMc>v3a6c47%~6byNbuRv9kEWPzlVtj^ADhi7Sd4s9&CE7)i^eeF{E&7 zGoK+AWE2ZSf(@nAbP|_Gn3;{d&SyQ)duy#e9?2a}A&QawiI_XSli#NeG(aN9--T1+ zIGI+(>OJ|lOv&+c`0+tMCq6J;AtYL{@e0m@Q4_0AJZz>HMO5=IMoVJ7q?lT4EoXoT zBDMJ&RrS$pEk!+*9t%7}a1@eBAg7m82(|Q1UZ1W3QkF|RdMZTCg;jU%P-*s2RHFHR z&_n!GNXPqQ6q@ARZ`TlD4<7FY#}7?A$fhdDFrqepB1)q+idkYY2C*0v z3iFW=Ike7BZ>81pqlSzB@C3K?D-w(&$?;u49eGX|5R!K#f=#;dX?;diWSh z8faRhqhi$qjk7bR=yL`j3_}UBuRta36Vk7#s%27SXXYD^IEo7(-q))3-Njk=FDiIr zYEVO|`&$hYC&iDnJaO`suJ7e%Klp!yuKs?h5}Aprbi}0ZHEG9=v?N(}!RV0{u_lz@ zvRI#p$&c&WK$GA~S<8ZvxVWuf3a;tfq0-%LnVZN~#+_+&{BfZ?NT%NUTu~=PkR_eY zwnF?iQek5pOmhJAqA`JacMo0&%W_2$=CjyrtP|3lcp<)n{e`P2tkDNd-d4I$TcQO|1CYnrH9p`$G0}73o#}6+OhO$ zXD0JF)Px$FE@9oh$tKk$$BG%|nzS6@Wu=)=;mt>;X%qtg3?lf7Ms~a@Z(SYXn?aPF zGAcvIXCMfBXsPRMrPHt;&{a6g9O~DRHF`Qh3&f?lN#;!b|_X6rBRyH5&i0rz_vw-D$z% z%4@TLAmIgG`brFt1Nk{K(q-_PA#Zh>^=S!&@xXi_i543+_KpmHLXAzTVb(Ec^*Me% zzqAtM>4sq@r^cVLtw25)m9J*$ zp>$(tD#N959&BpKS|t+xeZYq)h}G=wh4+K+RFDV==!vG8Z{xzEn5>u9(-}WsjYqaC zduMqxnf+N=xo59AGmPz7hINkJSe6gPASTlx@JZ%jEk%c+W=R>bvZh3xK@y5i$yvT< zCXB=ohR}B)4N=yDTcTy;mxU-N^upc3njDc-_{*V4Qetc9md5&N*G}P9#q^`>$pOsw z@+1jpOIOJdSeZ%`dr~g?V?Q$QI#e1pH@QP?wv+$XhHAJHpf`(=(y$4CDfG1&Arh(1 zp0Z&M#^syJ2|8(eY$&q4^OX4uDx(a{K(lYe6Z8yrULVE~&wCens zWc|RUfF%1Aqzc&tZa3jwooqyvA9>HP_U#8ShH3PTWtKxsj!CnQw1(M<(bn_u_R~mV zBu+}j_xI-l(4#D)@U6G8OjUhTXwaN>-cxrb`*vWVoa69L|YQ?6I%cGJb zCSZ@0l*MAmlQHy$$sYI$F*?(5R|uuma;iUH{%LRwJ8j;I%6gOln&wRPd^tAWrjZZwTf8HP(#h1 zKC$OPtIO}91WWfS`xbz?#h9$~QUKp6Mrg46*DnK6CxAw}8LsF3fU*ZlmDK?i*cpp0 z>AScg;i*z@04;B0jO;*AV;{qPW8e&a5a5g^E9j;>yPPJitxg`k9>4ofGs9(J7sTXfJ2i&FPn0!0nx*z3 z`wZ&rZcrE|1WYfaU2dP0*eS>#kZ*hMX8RI`3P!lGYUh?3ulx!rhmJXJLk@%^ z4HsZFa>nnq^My=O`Spuv0N960jz!CG@LQFXBxi5Td#7*QzPoMA_(`4&i)q*G7aufQ zlx{4%?$1;7KmjspF1{`}L$p=jmZ&XUBbxyyJWo`OLmoTTJPh#ovkrv7>f}^|zca=b z346ohp;kwFz(l_O40J6RUmVDB9jIsneGVw3I!v+{fFzHL#|ZVvV))m8w2;fFEiFp9 zWSL9w4{-82#l2d4EjQuf-_AN#>Yxzk#_E2(n!&LxKQfhbMsVoLLgiPgjnE2-a_a~A zoezFKqu&u!g2nJ@fDTMoLsWfnPYz8q=$_{Y)4RC12pglm$*OEd+p`m{hPJ&L+g|3l>+g*7V zAxAZXCnfE{{0cm`5?a>O+n)$XWh7WNA=(7h_15z8D-Rj^tqz5@W;ZZ%KGj-YCWwK` z1%y7xg%}l~*&_Pl+CWF>!mlAa^xseOmKd)RujNiD{+IS8)!a`p6w+2%a;7%g$K+Ff zU8E%s^6Di555QuFw3(IM1iH&SHzSE}Q0w}|<`$hv`|DlRr<(u?=%bGXpmtKwyEjn3Ih${gvX$D3%NeFRt40biNTIvff6rgWLO!N6>_&{q zMo){N;Gjn<(rrhx1W;V|YB?0;%`;TDUVs%f+mp7P02MzSiC%JdrC#K6-aiU4fN%y0 z=c+iTr7pe79lwY;qzpsA%%*<-p#ho1G*7~GliKf6j7bLJ4mu00B$S<7V5!R(HeVvE zV~=a|c9}OJbAZxfvZfTOW*fA^-4;fbY+_MS9i@q1rcR>hw#T72JIV$2uPfOv{@8{` zBUlgK2Nu0T+xO+ny_n35=$$qI27jtjv%03jR{pxTW{+{@yXpRM9^Rb*h|U0xQVeyu z?3Lnj&39Q^`CaJuTpTeVQ)JmyyP)dsffFK04!q#u1l;?*AT> zx8ooHkS=YCUxqJhvZE|*)zam^5uu+E&vs;80xeo|W@Xhh3`22{H9s}#yc-YMuFlFC zyH6Y)CU>Q#Vt7=aEbXT*f&`Twp-GrxTYvO|Al{8;1fhAl9%-wvGI2jlsD`Wb%V z6&6+&uFEpWkmgynEMsv~A375iVPOQ00Tk+iq3dmpNY{3QQWfL-Svl&wI6F|yI9Tp) zcY@S+9hs=ZAWxDS<^WO2f&pFYBC#=VXh6qKvoX?N(EC0#e>lvg&`y2na6$i<1aB)) zh4f>Cm{pB`kGN<(HixX;_UUA+T+X!^}@9poX>qwd52?yf0PS2 zI^i{z+AWf3;lZglI!UMjB6Qe?^;e@-YV+OjsKxQ>o!=7dt^}7K-co1vAW4Ur%9EH{ zVAm1Tr%s+U^O9clBWIZj$B;K!LxNqjCG1D+GUIy++e$XsQTlr1^BA-;53qN*YrjHp zj|%Wl*}xH_-{kj{l203Sk{r0 z%5FY_(j=~~a*QYjxQzO*F~!=fPdvdW&l+;pUHTF6X`WfqXj%PNlLr}Y3*N1|5oO0( z)6y3MzpTFpgXdo%1Kx~+TtNHqT&{1oO6@Tu4uOfahmUwrJff4p=B=|gEQj{KiY4*M zNGvSnG&QGsGl{p%(E1JR2<>02Ks3KM)&DxNE#J=wuJW8z zdW$8-dR%z^_fx}QPQjfRW9BTjJMjig1&+iTXHY20V{vSjKI2fnhs3%Cl=V$cSAw=Z zsX12|DyHuw0Y|xwO{$fqrj3EUPf%GLDRRY1Ij)A8Nq4@W;1C`nH}iy0qc?RM77Hq4(oZgQfIb)08Emo82EZms`hm00tezB z9WJEl&!f}pXOpz}MIL$NdL9>q`6EB|i0FayGrBq6-7d66Rj~=t6GU?<|305XDWU%* zxy#qt_4zc33)qLMS!;<2LqSIT=1V((b0@BzyO8&4i=jz7UYDY7hTwRLr+L=VU3sj! z1{~YBkq|$9faWn?(9v{1?3bW|+jqT-V`LihuCZH43T#Q~G0J~Nc)w24YHle1TJ>we z6v=tVq5AkmOgl1`gbXQVGBjm-?eD+Pf48Bfe!K$<(%J&vNFxUT;XuD zYlj9!WCdGbh_Vnt6-7f(#Imyfs=tw^t@3v2S6<5fK~uSKrul)-iKA{2O>{)?a3lKq zMtiRrGO)7c{dNfi{SQF}_f1fN{Fk5_f_Gs5zX+;01qQUFN|_y~{{mB0)J_PA#2_G0 zKoAgt{{mA<-H8rxS#RIG6!HJgO92rauHiusn(VL~y3mpT!&?3K?*C_%)J?;RktGn` z08WIUqWr%otN*zc_-lbU*y6+;DF9WFz8rD)VI`z0pN|QiSt=j%is+{>`onl9XD8{ieJu{KO*zjgMdMrkqw+6{-XU#Mqo)^C9<@%m7Aw1^dVie2SL z(gv<5Pb%zh*-H;Zj2d0dfe(Rxnvq@gLGL;O7y3C+DQJK1DH80?IEJCXd(_NuUKb{L zN%(U}#Gi#fG0M#5u~P>)VQrMZyBHOL zJ!qJa7X}P&g<;A(zC={j)~in^H7qDYe#`24H2gN2e&hyqgke*~z7x?Kth@<7O!7zj zP56ep)Ty%d$v=RlLWz>6+*Zh~B$Klv7E@SEvog$*#<=NTt|UL;A;QF^x*}&=jXlO0 zWkO|;4Ac|;2&weMXJzTu6A5jX(vQhD;Y!Sl5f9UyoBNX`hl8mpbjV_49vLU>8^k>6 zFI{A59?yY)9q*SchE4Ge&xt12Oe3eLe9|Bp1pKXCLWQN6f_OzRKwm{snsAcWhI6ZI`^eX*X%d95zN$y;`)orRweTm`mSEbr&U7^!-}mdGyCfma74H=vN8bq zwuATidKvD`;R(Ds+&_=k(WyNvLw!8?I7vHWj>^!j87s4T>m-zyfkA^=WkkRM_}5HI4F6xk}%{j)~+bSl0lvZU zBAXg8mquRISS^qP^8NO@&GUoo`Xr)8APJE7HO2?#zd(&VSmBtV+litwDF&$q+`f`shX#}iX>hdp>A zFkrHd5uxiq9|qp{<(aVVao#Q#;FwL#))|8fPYBI`AP$3<2!Y6OtR$O? z+;*nxD=auj^n@ZRJM2F3?+vnOVj%u)>POTYHiguOOZ9yShdg$3Ff2GT1Q5)`fGoN> zXMXi^2*DccwST`8D>r8?Lt5GTUOv%_b}6&?IC6iu<0A1<`d4b)*!{BBUr@9)0pUET zde9+oO8ZK@mp`VDuBNf{Rsk&dCwUVx%L=QwGg5NDKPEC{Qe6EBg>JTUjM_W%TQ9FO zGPf=v$4acMGFxR|Mx%z|6M)P<2}}fjb(oPUgS4UkCH{#_>mS;C#j0+i+}P53$M!AQ zxdlojG1iAsX)%Z$A>I+CzYgERUW~{HL}Mf=!UO)j;3xeM0dBja0K@|o*PT{5R6)G5 z2Xd#sD4wXW$kTQc{n9}{(i6U zRgoK2LheKUFm|B+R0RJZG?0MoAL5Wrk7lqj7ZY&S>-MDCDN_M!d>jhhY`%&Kcr^=G zE;m~j6;;XJyJFMF)KC=g$@cw?V#w0e`fOSoT+5tRO&evVq#hP`x&)2iJpZ&-@VDmFft+n$jdt9mNa zU4)o|`LlmYz3K(Gdmmq2(OX=(9D<5-x>Z`$)!f*w-YhS37y^=X0vw3R?fN7BH5aS| zlb|cb3YztRu;=~F8etMOs*N)@kkOyh@AmbB6NK?jDhO~O0#YbFh1okQf{+mrwlmp* zPbq<;nRndpDsyxZ{-0thq9HN^Twc$_b2H`?->NOjL_aad12?hdy2y6&8)+pnb~??hN}zz3JL@eAXwF=7!%AFGto{VP`T%%F@jq$IvxGhW=yE7_1YR znW8p599f{>W>Zsh@UGe|VL`QSK2wus*rbPTRLF+o#t0;Gmo5B{AQu~ZIL!KtT7#h# z-xOtm**hx_Clol4Uy)?|6FtV^7vn|r+D8iLYyP{WMT921?V6FADGduW3VU$x(Z3l| zq_U%b_N1YI!A;Mm(E~%ys}?1J((YjOw${nQ8SWdrtgn&xF7Qd#Mtk?C`*Uq|V^-z_ zWR~;737}29ladgPJT73v?(h?G^XBKmg1FSNBziXxh1AZ#<(uW-0+s>aP)O1p$!y}6 zUo+Tp4jHedw{W$p4zdf$G&xDlvOQYFYPof_ObE|UyTZ_&9=iY`@1=f&3KD4Pzt^DQ zbu)aguy+b{Lvo}f)T6CWy{3X8(MGc+Nco+qia8XUR2}+%&~;A1fd}2T&&0MTwrz7_ zn-eD!+mnB6OgOP^+qP}ncJ7>W>wI<3!+qUd{n%aI`?vO5j^IFrth1#KBx@McF9f|b z7@JzogI0b%aN*3gc*vp8hGZl^GUk2MYj5zⅆfnoRIzb)_T{LIp0c8LNy zGbL*(IUI=xwn5Wmi~^d`vavcN6N$?jPzWY^7&7;(PLZ{A4bQ(zTszBm&e{BroYz-Z z)zMOqqetNdsV)E?$~aQocMPabsE8%_JbISpQ}upihK(Ha5TOFF=mIv2UlOpOR>y){ zyq>6vb`5<2+XiQ`dR&CX^ph*|=$Sq9i+w1vSOykeIee$uKm_BWtvV6NadxR!_xl{* zxE}Q{B`QyN9h`3}w~jU+#jqxvw9i**DKzqJVU0f;2&n-a+3p%tms<4&s%ZWg)_(_? zTp{$AVO$s0WkiJAvvDSVTBYX$H%(dT{I*K(0BgR+NbSp@((yr9H_V#f*E`S0@HHss z0mA5&*})Yb=J_77HLd|q`)eC+|G6s6-P(8iDDXaW+?3A(W-GDPZL{H3%)$i)rj#Ow zTD#u_Tb@8~{$t_zq!?rJCS2lk((Lh#dyg&NuO}atdLV*w_xY{8v+l3twrNY_;;L;$ z-Nweo;ZDzI=FWn)m)8>Cn(amXMP0{ZrblC26QDTNy;T~MRA1P}YgD$dx#{Vj^V%c} z_hArYQaD;EJ)GU8#U|>^g4|7W9{ou~)HuR9!UeD|ls`)5@^QQ!34W&iOH^Xr+b;}s z^hA>}JsHk{_wmeW&2n055ZRDP_yE7C=-AMYU_4@-kZa^SsQj^kP2m|K_A1T${T_$8 z^-R9A^`+&J<;A7%LydE}mDe&CoM=J?Ui6}^(Q`hdA|V>=SBhI)fSr1Rqwf$%07`?! zLnP30)_PaZh7FC78hY^U(Eca$feQa0>fUTicrSGI8m-6flY+Z|3-z;Yq4Q2(Am^U$ zXD`#N)Lc!cK-}0uh@)Ejl{I~k3&+f)OnW@#Ou6RP zS#>yWFR-n`MXW$3@mEoD4NFMZ()P^Sne@W=`nXI&UH(}QlNEcJo|#mH{&Bidl`w$f zde{xF_r{~YG*-{TSvn4S$C&c6v4-?k-8fJa{1W8>Pg{EMxrj?i^6)I#X6k1&zQb-&jIP zsAMSRFC<~TevaAB zQ@w!~a}GqfGu(II_KN*q^&{_YdR4Z4j%lz)V-NHyDD6WG(!*7H7wO+_DK#wD5qKZf z2|eON3kV(is%=q`_-0}2;O8KnPEjvNF`*M9OPBSRZ`M{;1-CmoB!oOV_m1GlCgi-m zy(6)*`;{1!AY*7KWPVNUvH@bptl4-X_V($v>W&SXTAMO_g7nnVGcCKMN#@$Ktq^Y^ zZ-z2-9TJBSs%7QgaKCSU`-d*kyZ!Dp%b5&oslKppaBM-W`VvQbgNa|IsKQi-QW^1Y z{p_Xb+(7kOd*6;tAB&^WKRm6QZzowXFKxvy6ji@vj>@uVuHPjLeg;~SE($IQR~1O; za-U7rwEvBBe9-FnR#x*=WOQJS7X$|-@)I@)eQxPodG#K7Ra8h=Zf52^B`yf)ywJ}6 zW?1giyj$p6xBRuSh3MA%v8d7+r?T+zsJj7mb^S{qwq-)M_zAQ$Y8XDeBRi?{dIa<8 z!nUF5-P|(7sWD z-N^Oodv#9uTmg`eQ5`N(B?PM8sr~zr zs_ImO5W*&uh(uDe)2%f76_Y@R#|Rjx#}B-Jly5Ii_H;68PnCLSgfRk!obgyDU*&0=0)ZOEtDa@$vR%&=O0*S7a3D)Gh-`1@l_V46jY`_7A^B|Xf5Z)}>{rxvfR@v{w)zo37;%g;hXapKOS{(4T zg8U89gV}M~LU}@jJNAS`y~0k3C*tj2yGKbn?F~&|0Nz>`-guWueij((=tn^h!O%O{ zNmft<0P^*vv3=D5IAmL+R@YK^F2w* z;0j2njJ392J|%g2`1?9-H>5|9Q>I|-Mw)~Gj&v~;(BC)g4T7j3Ag+%xC`vb3sIV)C zFI4%v6F_(F8S}xoOgpHGFxa;gfEAz9%vmjaYml+@I!CK`e?8xGxx4Rsc{2Yiubz|! zF*GjsJo=ElTcn31*s~u0EG6lscaEX9hV#Z?93`=ps9?*5U*|me3<~9wpQB@EY8h=z zVD~d?#;bmcbY|MRiFhl_!vrmN@E>}@ zq*%{T7khX&C%ChChohP|!D;_sSxB}epWRScUvt+|BTNk6r?{%~jNeZ`*i)VSV=s{1 zql<0a9+JgHD-a{8W7i*`vpKXIm05ugW)Je3G~RC=;{BFQpQ-Gh250UTZ6AK+H&0=6 z1B|IW?+g>?BLAifg}SdP+|32E<_1T*$7|iUtgk^$K&n;^9cj1*(-HQB|&g9B@F#?#w`=p9k zi(l#=xo6GNj%??KsKD0Hi=2NZjShG=+5oi0No9Wy()`jgW!VtxC$w~lq!%p|c=R6g zCpNG9XP48W(f2~CM>V2TNZ%fA-rd} z>TaWkk`{L#lQk4kJD#WYs8Pfn$7&B!o0%zzuTc0>9$8t01tNv&ovwxu^oq|gvi*Hq zv>Q~LvsSSfmwLII^*pLj=(%oH6!@+jWcdzWL%M9chN+R?-6+bOa!ZVEaUiWyn^E}r zxeVkbDd=h=u;CikSHSgpc1+gww0WxpGnk-B z724&>8+b}muEvSc(HuwkaY9jGzY0;_nrFAq#8J0=+?y{r@|(2OkV*nO&q++$2;|54 z>9&5~Vq&1BDPEFmtA8%009*7}`uk9my7hlwFqqSk1Zrs3*5PZSWGSEUt~sMFkw+L1 z$+jyM!}sD;6@yPj|I&8DyA+H3?IZ8^oP$UtL?hKep#Loao2i}l%dCjS!gg$$?~BI1 z=1?(G=jixwpdt2yALF9U3$@+~d@M}gYmy55DYxksPk zD^zYCkVM`=ftL}Y3<&@>(^k&Z`z*N5piJtjJ_4!D_A1}a`8Xv+B~efa$^1nCz!f5hM*n|4wMngCPu3*8ulDjgoJmrk56XnfWlHU{tH&1olOuf`E&8josxtV=rZJI@L_N8cGJu3E0% z8w-R<;RTOBa0Hz4?mYV*o!RNkYA~O}f7wL6Pq(lOo^``@*@Tg0yuk)XGF+;hPsl2+P|0!i1mr625&zZ_w%t-@qo#}up9?a8 z?R**)UlXwJ7IcAnu8ZyKf&O%Pk@Jd^|3J+;U);)gdX+dbJ99^!SjU_=pua{W`*-uZ z*99GaLVr#B`IH-RonPWVA|WCYRAI|2?C7n*1fcgfvClB1OECw;&xrUXQlr2=-CGIw zUD!AbqM^QbxX+;e+=|}bM$i4)n{>tH+2}E7;sKPQvEMrU+0mJO@&|dU12o{jiX|xz zs=us?-)sqtUvb`5E_^}wZIrL*;kS5s*1yQt7t0_M(= zC_u7&g``T2VkPGtD(h)(J5x;6@S8}Vyiq>d!G|;n?!;!3k6LBtP?!U zx_nmBH2sC++5MM*+IkI&s^vo+re$N6eW21+(ZccGlJjvS*^j1dBE9dGd>Bh4qKqFw zf<~E{npY8nEhP*cGI4^Sa@9J+*tDozU^1r1AoZnW@|Qkn>utRt0v(UE0f_-oa+ho( zraSi4sTr29dd#?CMI6)ZFe}WDa#Rol3j77R=i*D1>M|OZfatT98+l53_o8WbFaTjY zk2PZU+x5uxLHu7S==d%%<7<=o+!r1#T%C9D$GUypcbziwPn3EC`7s`p?QDolg1?7_ z!W{S|*wdaDQi4hc8r`K**j@5_<+saV87+L+?};)lMej=;qbeR#3P__0ybl&(>Sq(6 zpPWF)Gu`AQBjW%P*7elW(eB*K18}fbcjvZyn%CUIgZ8s>tD}S4F2YS?cW)u;d1r25 zqUHH*;jhB(;k%$+|G*F`yl+x9s^q&` zF(OsR!Ys zvmR&+fxJu6`^^HS?Kn)54I2jj1U8o`2Yu2OR)Gh%oS>J2@n^g($!fo4`U|)qu_}AB z4$s8(w?8nFn9VG%69TGCIY0_TZSzd5v^gwCD1l6_cNIDbu4ACX?4~S1BJRai%jS*Q zZSxP3rh!~WUCv80(Awk>AtIgbxWGc${R~uO84D;XJ_y(VANeC$4cE-kT$mahv1l5K zRORwBrS2=!Z{Ky9!7ku$S|qy3w=t@(Yz%jdLqzxx5Sa(2i4?dkNdYdsI!E6lQ|53P zzxAnR>q&}6Fp!?VJzQXhC4z{Mgcw}NFS#qq4)E|8wuQ(oC5MWH*qJZkwcqwz8N6cT zP#Mm~)@}j^m}X;t^U@u`s9HIs+1v=Qxtis^7sD9y3BfYe@B@REA2 zAnrMscjetlUFVfuQUl5B8;ZJKly8%rZnZhhkBebv%5;w)U5Z=*8p#WxiOxqPJ*450 zB{UU(O5PeMIJEIg+bm8*uAO*Z_L8RNlD#KNi4KosA`eBB(hX`wMK+}nTKhGE>bg_# zxSYcrBX_~5)q~#SfgTc~k-}yA16My6Epm;Se!OLPuF$V1dms-_Hx`)KlQFVt_*grG z!(>7R9;(EYGbMXjsqhUjO%ZNny`&4ZA``~BEnh5R7#yxKA`#G&4`=WHi+dHD>9f_b zP1cw3oO_yD*Ynhbo$C|>+5^`T1##Bu*|>`xVl*e+PSHU6DT~>ysjVYX(|}WLTTgrc zbp1g!Cm;pC5(9n`U$rjgP_Pi`%qmJhGujADQGh42S+J+5R)3$CU!5qf_Sfd?i4+Sz zTlS53+0U)fA4p9vqosQFi^);KUkPZl62UPPh59XIcSmw!N%d7+18D%rgMD$_2tt=3{kY+b7v@G|&?Q@uin09n* zs7?k7$cVwLe^%yr;F*NyEs*&mxw<9O^o>R2eSebPD@2Tv<|t(g<-hQ>r!_hZ`yO7N zBN93)-~HrKJL}XFTDD=#{nZ>4>SG45U3hH3E3v+572{%$4c|JL=f3ht(YE%|1Yk-c z5W4b6T!-XCN63+Nvrem|`*y$QYRdwLH|Z(@h-Jr(s44Aa9n3u3fu4c=lcoe1>5hX5 ziG8JfC#Cu2mOdu9U8_$N{-a z#hXi^FD*MNPREE&VaqtTL~s&I{ke~9%LzKJk_t9!B@U?Su?;u_>#^7oSooQerL zX&=pb3P^oO6llXw)gZAV-=bcaVdC0Mo-qJGg7Y2}ZZmB{gZ_iv>JBhNgYH4BLvr#o z9_?vi(|5C&M?vdSbjw_}wu`cqO0MUgrg?c4cw1buk+Mqg2+i5%)K=L+V!m(CMf177 zbE^EeU-)Rns_Ox;`6Bbijp^Tju4LrT%WM$vlKK;-qb z+WC(5kDs5PzumCNnNj@|T!dRElzFk3d6n9KHDC?9^3?yT2IewjQ9X;iZQLI$0hgnEEg!QFhu+rvespidaI2 zC7h^Sx4$7J>wDU%?Dz>An;(`gNuL|7JO$ntBUuTuLORgg>FwpTX~9Mxxbx()xnsY8 z)8zjK0kCb$Wx`O@kUu&-&NG?yE&d6J=XY_etwGVzNMFWmu!(6PpJa!u8qQ}Q&Bo2f zy;mq-rZO?G?t zN3$TIs+-+~CqQd-6B3?rg9?Qy)z5+(d7y}WMnh8A0~fHwJv<=W(FXvYn&7_6zK=R6Sgl9 z*Jzr--s`IfG}_EqRcCNAz*V`$^da2YetfX^T;8FUuL0RtFsE@chYvFdXYmYZrDp6RqExEI+rssx4}?5&R|Fb}JJ9 zF-w&G(e8~2^v97f9)~ul?}IB=RwHL?z=THIE%Gm;Hf??J92(^3jBm%^4=OAD$`uo5 zH_dXQlO&`T58>5yM+&$mg}B}3!;(Fv5_+)LC`uWI%+aInJ85T6mPkWlW-42CYw0mQq} zxgSBH;=XO!cTO^j_Wfe_VubbT_qc`mck35kCl*^9TUR4!Pu!l>Nd{>d+*j^ZRp}p5 zxZTJ41Mwsu2Lx@cVb=01=1T(=FkDvK_@qtAR;$j!hBdSi>g+c&E6iGQCsSx2v+(@t z9ky4)9+;vN3|B+j^gYB7Wgvo!fOfg^?N2CdE$ifMZk|nzjgP3+!X+y0np1tb4g0w% z9+g6T;wn&#Qa`(KL;N5s^(jS!jNQi#*KhWJc}y+Rfx;+2*$cTgttFGU`GJnA0EJS8f=Kb!PXuoFgyKoa61UNpu+rT-k zaOkCWr7~T|wHnn|-F=1a!^=?gf_n!S25quvzX}I8Nc7ggs5gIAmjF>yuA})}Y-yZ> zVzTN+KE0C>q&jcNnnz>L6FoTaR`(M6-SytH2T6GLTR@c^Lc2PDjq?c zoQt^1G}TGgudGf~Y^#Pm*(olIA1}k(&klk%qKrN4kMa1&c$B1+fVSO+=f($+r8u2a zr^=(vCQIwS`Er&ay1Ehz!i$Hej!!h$&@A>5(ze!pTr^rHFO5%Ep-o)Nww8t8fm)Cm zwff=0Te~XlCf(e~#uIE?9-^OX!ODtO(keH(eUhK&wn zM{3OS8^%qz*uXm>Amu|c4~hrv=Q*Yk zivhmIQcn&gAz58${%dn@Za+rtWr~7PyBJHSrfBm6UYG$F(kcDE9J;7hM!o!DRPMq_ z{qyHZUV^o$Uy*_W4hIfhZ{R9A?#ouaG&P5~uV&CsPd_qwfmQUK2R+2H%C03jg9Sd7 z!(H*2Ic&_IMkQg$dDi<@65b#BkWGa_aH#?B0==Wn6I@Fd#y0ArrezFhdE!re*+Kdu0*l+j}w;))c_lJ2}tQ! z)IFQiz3wQA1#WxlvAJzl?*nD6($sa9&uEl!VlaGg2Br$PE0LzITuRr88~mg@JEN2Z z&JkCXr3~G~MJ>z=yIh+&(TP(_Q{+eG=GvznYs*yMQU{W|7&)U4vIb&p3yRXGUp+gf zuhf}Bd3RQ4@F*1}{YLZSf7+i8ORfovx22DGs@ySQ0uR#H{CM4NE^}JTjt{C3Fi()F z!E{chd8xgrqdMYIS(5y8-A?qW!|$QdMU*MmzYiXC?e4p|y=8N~4;*Om+yR#mlqhOd z*Uh{}WX04xb5jeUTNS7n1YiN@y3}~nDwlb+ z*^GzmJt_DeMM3^1Igj{dQaA)5UK%v`BM&+&sp!xn$D89i7f-rP_o;w{Ot1ZKpn+FcEiPR&+$Zz?r+RAe(_=k}p%vTo^qz)s#`rz^Y_8`qXX^Q5_ z{S?-mx3BHQe#sl%4=x}ux11)Yy!K`9=5ZguAi6g=BEm^@n3i5TJ8<;4*2`sDygzgF=A3A`(N(FfCe+y z`_;e4&t9vVz<&g>S2unSt2#}y3+&qCTq4{?u>YbfLI z!1q_g_H2t>cO8DCAuO^LX=5REkNl`A z#hgtVcFZJ9gEy&>Noe)7JbjX=chhsNwnr4E%X*!>OVp*}XI$N4RXix2|I!{Kw)y&QTY9JVyl)k{AT z_m+K319<6!p#GM4k z$VSUl3hR;g5pPj2Qy}6A7}y1;xdcbITkuYI7c|nWrH5_;8OpaMkmlW!RlUpRuKty~ zB;t%D+JW_Q+8vco3V8@<7L}6CI)tRSMNVS{0@S6eCa4`mmQgVv<@uN7Ow{B{%+@;& zGj{M>oJh|~IuvqQuA0n)X(6Mbz4eeUB|Iw;YxtF6R#_ZrvwK4-;xo)b4t+y5IqWjH z@%DoFJya-EvqSLdp$pQZM>&=jCd)g7y`i+BaPgklNg)-F(o^@-au=f|as1C3W#4LI zgFZ=+6~yoQQ8Z5g7zbyKq4^z~QdLCQos4EPZf|hFq3>B>iumla9S?^_2ly#}HSG`P zT<3~eA^+Rq9FDNx^*@*@{P%pMj21SaIP;>r^}wiEdq?Q{^oi<2UV0;2eZfE}Ow2YT zrR3s^Jme)udbL^uctP}Y@Oz1x(F#`MGs9GJO0?M6L|o7T@NilJh{9ZTW37D(<{YR* z5`v>pi0dYbXi}xIK`xvW(ew*s7B#2M4mJG6#^zRzdv=6a!;Nb&A*+L=$Q6fS$ z^l2#TK~q8PMgf5f^Jx8CY7ty*aIA7QA>VbUUG7hYTg}^+;CMTPkC6Rj_}vk+%Ys)^ zTEO1eoEPK(8!|Y{Kp$c9oz&n^(l$gUZ&D>ob>%06i%5$FzOEN{f0c@Q?xQ`JDdgXZ zdG9y-b33it^H;8^8H>13_APf+d(oi{QI))05h6$Hhs{u-k>?hB)poeNPBW>1Qd#2q&a(N z5d0{Gw+X8WUW(>lGuD|%6^kuWhZQ`nyGu_SM(YMd*t{eU?2`EI=3$kXh=>$h=fuBm z!M6mYbh-E~0h0zB96y^TIa5a7d(R(ml-#_|bc83knw8N3 z7$~ZFgGyF6u;92?M*7MG6IL^^Zf zIMGBv&xHXa-H?DX04eZ&LIjSJ~qPst$*)@ZNUP5fp`94 z;9hX@(d=1*@V3nrKd)i?Lu^EM?D_OSQ}#2+T^Ehl&LzyGMO4g5jXnrpn)Pq^iqe#a z97TI!*Vv$Bm-trMYO*A9zEr2ht*$ON;|8hnjBo&fUzFWX^iDzS9~t5G4ymV!{ggNAZYUPLjdSDjW!&(hkgJ(t0bX0H%9`Sw^ z#X>cB7(=Ol2+@I!w&~;h)qq_I>IH%4>3*M19lun(+|ZUvO8VCOl6qrEX<*`UcHX8$A!YX7bzzDt;ly}*QX8<7&WAB~pQ2o#M%7K}OW=X+V zP3Sx~Pt5BJ#-WKqI0hSMfEOI|M?hZnFjNlBkkzw7Tq_q&!NFU!s-I?c9{wOMu*D=+S;@`dj-Ms^)G5F)cc|NX1 zhPA^KGI&X)vBpmD!~(Z~dgWx+d7oA;MYLr4Lvl9K4~nT_B3Q5@bh7JOiEejwXu0MY$F%A%9=af-T_K7IAP%Nw4 zP2{TuG%=Qt_htU)pZxgjd#V=q1vm*z=Dxxs@ruhEG5815{!4CDR+i%>&+)qfEqCq^ zy*vR?yRIV>o2%c`v}K!-=4@c^v%H_yTjq-B^V#ulini!1mGbS!9AF)2sDiH}svltH z(Wx<-av!w-T^&V99GddRiJTI4x)0w-@d{;wvV{E9q$Mqq!gbHW4UNG2FL8VM*&(!a zw6UZT)Ko6)w4g(>3d~gH4XUb36tN@Mj8%PW`PtuKvl~8LS1xYZgnYm+%O4!dSZ)9%)(orHFW1mX#7)}OFVaXV2THG;EcYv=h&W|S({R+W8V zMgKL>Z3oS&1oe`)qT1JnB}5~k6Y`?v zepr*(rHXVB=({zLnUC8Sdf?m6NveYil~jAkGQOffAS9(n zbhe<0K=w9RBX8psPK~RCR#-fwB#p`g;b?}pcJ|#B%4b=74{?KT$gn5Iact(8lW9kA z0&3cJVu-$OCKrfjoMlM}&s^a6Nb^_PV!reqgepkm4(_JG@&Moa%0B4JFouHsZ0ZL1 zWE5L9MOE1g60-AWFm2g|Q^w6aG`G@GA4}ly4r{>nP=zqVudw}*Z^Fq|+5Wcwp+4J2 zFBtS)_L8h;p~x{q&T1~AtwwVusUo}_=v&$fm;23mBN(4AZ~fUo01j#IV}%4UN>7&& z1ES`d!A7k4_&f4TPvt!`w{{8d8LQ)?yX>Q4 z_=LW4HhkT&ktDeS^<7%Ktj)*EqeYUN|M;6_dO~07_q=+URX_zgFrxZZc*sof=FH3?z!21_Q6vGdtY1Nz>~o8N&J~M@3a2 zd$l~Wn+h@%vM!qaA3h085itmWuOeJ`Jlf{&{0+ZqO~0-r8Ub~Et$6dXS9s0I4dfO7 z#5-^D_Xsv8Wu28|_z>Ss>xBN63zEw5TtDFshu|$UrP0YR!%#$lWq0OrDD4c6mb7596;u zN}&vm5OK8-*i+qa<{Kiwu>>+}7rQ&cMGDUKHYQb#K7VHQvV*QAi?QTlatN9=1GG84 zxADG>L>F<2?H_jA%dMKLUj$Q>ZR zQ!lNwD<{jg7$^gw3DnAu@;R^}zLjaMrEEQn;sAKhcD~ z5Wa7*Sf%3PodL?W52S?hhGp#cxkJ?)JFsJa$DJ@A2a|uZMXeot+d6&VKj>Pr;`G`< z^bC#0B8LM}To1b^X1-G#j*OFZ{d4-uB#P79o=R8`v+%i;TF9NA&g|P^Ax&j*hstY3 zIOc*t977V6=CJAj;LtURCYQpEOp9i=kiM4BJyi7mHA zemq2WZrrlA2#4WbTeRTsrD^bY>uLW^H0j2NSPrHqv*IWW;Hp^e7p(;DDeTwFkl=*k zci9ramq!oVi*^($9^9zY$FfWjHL zVJ zvxgLI5+=`PZBj(2?yoGp_kUh3;qdvTXY;o@nkDQ~2RJ2)ouGm2QvLE^))A>S;Qu_( zzWwTez|u=+M!1j@_OC2MezBr7-30h0|CzAF9&zM-fKG*cmmjz?@Afp2-Tw*;1pKf} zC+GD;2NThBH=fufw2}vEo!6)xgzw^+Mp zZ`?K0%=C7Zr?|X#&DsCL=&WKx`4ibSUZXWxb?tUWyGdwv1AqCB#4Y;T>BcB%p&!W0 zYW#-;pVUk!>Pm%{3FP|A3LY z`Q|F6C84K6w$NV5Oz0Uw6mH!C4jIB2kqjesd=PKeEN6xJh&$io`Z>cFXh^z7vPu=B zru(rpT-D{tpNQUwLe6z!02zDiaw9#Zd3tIpCp8&nZ2 z4&y)NK7#K=CIkc^vM*%uS1c1UG3FPoh5`6*d+v9Kr?(4kS3fVB@~;xz-FwN?l^~_q zrEkb+)t){##t6^sj0lk{9%wKAc$~IBA8w?17R@;87vGF}Ii##_g_IDsp%nz|VIRc5 zCxYD;(v8Vm5SyNQ($1a-UiFX~yaiDo<^L7&qsT>s!2j$Q9KO3ZJrYYw@-YS{(Pz_8 z@lGGx?uc4W{88p1LLXuzUr3OFblOz;;*medDYekl@y`b2wCjJnu`n0XkYr89G=RbW zKL#uRTP*vX-bwX8@>f@SCkDd*7RwgNI??`5`9B+=|4SH40w6K`eUo*P{-5Q)vT{zu zKpY5?1DV2cHeY0{ScO=Wxrifj&?V@!Szg-)l+f}H^jE0I%F@TLPx3cYbu~gEPNE*4 z)#YkQv<(*%Q&S%gRZO4H_+_VrD@w1wyK`BA&p7|ypL(s4=$V%jCf--U5e8KgSA2g7 zGWF7~4Sg5^0> zglSY5;RVv9`t@&?tjPC-@>R%xu-v4J1scba>ocKz{pmdFyaCn7Hx!5_;6PzPL750# zoRu#)97Q6vZGhf;t081)-3^cy$Gia_@U7}lqT4ID2CeDm34}e=$Ag@2peMtX1yr+* zQfS~W2OY=nRFi7u^Q313u^0x1;Vhho{z~me@k=}z(6qVmHPcs2RhsP-%z2is8RZbX ze41`Z4(U1YNA&)?4EE`66SIHIbc1WFqnA-3;wO^z%0%c>8mR1hQBRq#Qm*&Fwag>zaKCY!s(GWR#}S#st8k^ zVM{YOefZrh~EZz9JQCNK@REggPu;5EO&~(t{eb#yS1G_C|GGy63c(vVUp)<0oNBZdT6^!q3LN3pB zgbx+`w*Mqawk|l zp%;1G7dJxfisS&%`HALh7)@84A1lIM(XVthf|#83vnaq^|2kE-yNEKe=vG(|KOyV! zbObeHl9Tp+>+|9@z}?~Dwc+^^w=45!*ptSUzMz-GgYtSiMiVDv|BxGF)`q^zz}WYj z$3u1Rr_a|%>G!k4Jue5ZhbbM0k484g!L6AVF0c1yVD$?&Y7DzDUr(csO+fCql783Z`hDdjm+-mrL*7YH^c|GXNDcJ zLgogBvuZ;0VY*9hW1%EAY>l!Q8;j>5q@aL9DWBM3f{no?Pt>RRt@pzvL=Ts8!!tnd zrf~2N*&Nf|j){@-EY{^gAS(#edToDNUthn8Ju{G$JSY*^(=VD~Tnhyvb&UoRmq(c3ma$#?cRV`WHL#%DDk$I1I%I@U0}1RD5vFCsVKJI3446E&GBp;%0GM23i+?)mImEjrO~L_`y_(G|XrxdIO$?(y zqT|W5(77QQNJsPiuJvVhbGHR=qJ;;X{(eCImnd@e9U8P{lk7W&+3dgQIJt*Rhx5{> zTqaqwFl3Gn{}S0TS{H-af;|&9I9KwmCaU~DAi+vVp%1=7)?<%E(~1;;PRIj| z*(=h{DodsIT$(>&!3cvpqj$`a_hIN86(=}gg0ou2K~c$%e;&IpLupvu)L9)`ib#Uy z*jHomXw6ixCw_@YiqF{pHUo{LWj%-N8Rm&j3=GXc?a0EWF2 zNVDfQU;_>1RJzL%6;B0rD^%G}y6g{BCtio+6gcmKqlyw$#8!pdLzJ(ygK{O9;lhNP z`q3J@i8H_%;pnNhDoRsYqaMVxloa+?2UHt#dhp>4tGGCLwx{vg)>OS2v4>Vpd_!s1 zatkzp(}}bHOP#|&In=#_!v$|5_cyhQ$C@^R;+hyv_gL^Ifk3FTD0F&ID;pp9IT__L z@Q@5e8HnUOZ99}Z1kSol>?>leyOUmzQg0=k?9ZR66l|J4K*j*!5Q+N2aS%rL6EYsY zSGfu?m|*#a363gM-60YIgW(Tk{{$bty?uCE*QK&aZl?&8}t|P9Qcq$r>>)vO3xp zL4VjeiXLsgUqI|#-w~j@ozSv;$N0b)m1#5xJ4O3M}>1T4!Hc& zcOmjDiO7>O4*s_HtlbKHI3odhjZ9zv3^@R1QKHoQS~0h;_NR$>GNXk51Cc;(ztO~6 z3eB|ie)J9LAiN#QSCc^mSBqRozN2;@t?Kd)K0@0-!8Dh;muw=Q63ld;fAmIB1tO#z z_l;sa#IMpLDW=#UPRKTi>m-9W*3*Gxrj1rFHYvdl;ZLV%uG6|}i+_1_a(aGc7N4gR ztI2jvvcwNWV#jFm?0E3_Tsk`fNPioEl(V{ZZCb6G(5uzD6wiu4$~EBuymucTNXuiF z9QD-Xws=Qb4clxrW|7n;f8w5JpG9H7N%ACfrlgD>{$|VQFWF9#h2^*fItvC&Ll>Js zj(eNaP{P0jnUg3Z#m3y6mM7Q5WW?H~EEzudaFJ8{LAyC{yj_qFuMBSBb;{9$4otL% zuVbP}lIAbb$(?Et(4h-EO(6gxRzGH5%zl;`TABcL&Da-_06A#Mf0~<`2hYF#`rmjP z!B2CV1I||KWo@FCcI#E{Qb-OvDY(|YG!f{_C4$||q*sghcwFaoBIH*S`o!@U)^5W` z1I`*b5p8ztgRZD2%pMf2PurA5!`z6>;n!MYVNQoA@+#5+KzYr0a;D0MN<|abOF+U> zxec7WrE#^GTW^fie}n{D$K?&jeXgS9rQ&nYn2*Rb!EgwF%p~@j)-l&~TT5$C{p>+3 zteZYai#2CPndh^_u>IhyL_Tv{CkjYLwRGzZT|0s;C2Tco$ZBdO;jI7HmvNLQIb}4!|C|M3})$dD>Ae}-gya2OScS` zHshD)+iFj~e<|VJYUwnX76-Jnhj5yaQ%UJ46_if3nT&!2X3eMJeAcBVcii{jaJb1y z%|G<2)_uPi;jp_(rr9kb^n)hT>g#TVn%=*QR5Qjpv0B|VQ++Q>VP|sk5$XvN2e0#};L@BE zF*xWv?&>I;cEdtV(t}00OT!)y)B||l>A$QesqR+Z+>wyn8JFPBPHHzW(Xr8Cy_&w0 zvBX;7GB6Ka*E2a78BbU<4d@v4BcICh&Bi%$rRt?H7t|UqOg<*XBUH!0NY3sWbFpoh zN~uZmf68o%m{Y2pZ$-pZ8peQh1Mu8rsYRX@ETuKXDy96Y^0n{8*=&&#R3|)}AJR#{ zt$r|*OU9gBvP@1Ez6jnPDCtlpD}ogkZ{bZq+wPe@*hPl0d9fG}W0|IH{d9>~J=RvijD& zz{YkQnKV#%^}tc>r4H?f#R>Brotxe<QhcQNs1kFFa*^EIE9FCTrd@#)3L-6f9P5=Dxl;%7J7#_m9#4lZ`e>GrqhbT?KS{eK2$kS8%rb`B6_ zl3sHdACk^5&Apgz{qBW$@>D#~uYUJkN`67Tj;)WAL4DXY`po)+F# zAGaIP=SSzK=STgWiQcBD`=Yiiuf7yHeMCvyn2Q8VCt2yXQo~vf85`?yDPSMLPe|xU z?UNr)IP?k2bv0j?RKySf3NqMLKmM@RN}uGxE&%|-B3F$hC<)+mO~@etDHOiVe_>3* zc-d;{P~P_dO0Cv^7VLYYW2@Rp`dx9wr#3}thECljmC?BL=^ISJt~dEm`gBKfc@~?q zJU^7sp*}_HfD8EDpWae7H$Ra1>vpE`FkqBfSdoj>AqTNAf4RYD9#9QAIbJ#*fc%1W zGo=%6yxvyp&CScJ*W%{8w^#46e>gV|qYeprEvrtu;)P+yDr=b)u3KUv9`LLMp}T3> zU~np(6G!c;W!u1EUu-I3t?P6{Nlr@C&nwA^BCFAcypld9n)$LRGlW1eFSwyH3eJ=y zRUfD^zod)eIpUxcrgoBjN?ZCS-=eb)ZTRo^Ha}P$!WPueD9@~9LVjB_e-y0hp^mm> z6d|l!rL~>wjZJkXMmr5p@S-TnYS2um|3{Wo#1mI<-o1Wyd(~>S?Fhr?b>88yNyn|= zegp#E=9yp?Ti$-1JvAG%o1{AZBMFysr$RSV?z}}(YwM+Jk2tp&ydsmm||Qh{|JyyA{WpvX`YaO;rJGX)um zz9QIAPsDI$5-hr4w${_{*o$|q*40BwN44Tx3YTdFzJUSY1A-lx`xu8>0I8bO_aIP7 zs4;Y*Oxn^EgK&bhIL!f$6U;+s0sC-^?d=Y(V~oOuipQA`Uw_4!lKYZZ{F@NC$QGI` zpE;E&oovH0arTd`3TB()y>ALz?ZG^iIO{wZ^QBto8t^e#t2HbVo|fyO69L~2E>V4wo`c`lkfP&ik%xTV1LciwMTiannovq=A076VgxXm zS>o|Is=Z>ffPY?xi|sDUSE1+GiF*#GdqW0lJ9--NeVB}RpCKvuvlbRh31-cW9pmxQ z$T?L<&Mx{q*pkv~yfwBPsiULwqqB2&Cs%Q>J;S&Cwpw!%f$ILp%i{c&&z}+T0)og% zYdn2zaME<8qSP=3Hw$i1OeGznmDghk&xT4ysS1Nz>wm~m`7vvdb~U$LmMFQjB||=v zq`{%_8}lekBNSECQTKW<1 zxwGvK4u?5P?sNpYI^ZF&I`=_~9kI|w{2`R#*XWWi^MLaYcELs`BS-bmHwgK55UFL@ zj`xuI{bsY##U^q#>D<;Fd)L9Ez+*qO}`r?!JXop8b5t7rl@DkNsooL*1azY%`?PcqfWV-a`}RGAb+%DaHejNEE~293W@-E}H86Ml z7GcNrW^^Z!Yb44J`Km#k!Sr$g3~`T8%iL-tHh;m%?|)aBtJ4ej*m*n{{T_40l33ti ze%8Kngl%-k%D^4(f1Px}R|-~x^6}@r*E+*SHxTnJC9#Hxeh&Cw`rEJnBW}x5c_|+K zT%EA1rS?w0|K*V=dUCk}ydmg9g7vTTAj#MT?cHc=M$=y5^s6Cg{B?|EscZ&?vFz6e)9pRS8)GGsf79> z(7*A$J#7HU%^5!2F@uKM7ks#b#0+T`IzQpj8NW5oA6T{@>Prm6Wu^0Sy|Y=LI4FRh z7i4ORjdOkY2PXo~#%}-Ybo{#tjtpO*Dps4ibfk8$$)y2T9=zn~aml|}t_|~P(|-ux z7d?3^(66GxJJL`y(q=Mt{H_9Z?4F#Q?NmB(zQq&zk}qo7mf&z=rE7c}M=-G*CHtxo z4R{shCPjE-f($oQ$^UkDWxH)#Y4|=*L8!nWHY`OFWmy_T19+UYV<1g~G?)c;AW|gd za3YJUC_5t*0s0!fo7;JfK1!ct&VRq0jXb1eN^>*42x^LF;aUFu|NiCgzx=vk>2$_nLAp&YS76p0ODPs~}`n9+<@CZi)w zPo|jjl*pyc)BFVL8uG;B-Z#HL9!6m_JmkL7@=w5K-x0f`5%zwU@pfDi2X5vg6{F7DWbSdL344|+gzguK?fPBKeUN2)1 zJGoml>cuiD+vo&>I*Kw|_rsZ$bZ^+v0G%cp!VEG z#wAerGZeW(v8!Mrk|Y(d=kVNpfn;N2W6@s>iJBWh@s~yt1rHPf~I}Dj;N1x^A zk$X_yL!6aqC07fxL})wodyrQMWW#VJIZ_gn3Pgs;$bYcHADLFqX2US<4i7SGmB^CQ z7XElYMH3buGNJYsGQSG}fAn-hX0HOpmgj78+cNqr1r1u~VN$D|+YQYvE4DL#Gns`u zT(ht;N`P@K4_hMgz1K$B6DwDT-#F&_?D?S#it7AMw01t>Gs!wy^Og9*!5?`r=l6TV zFzNb@0)I)z`*VI9G`b=(3K)ctXEMcbDwwxb-T)XV0X_sLt|WTVO-4X@Dx4F`MMxM{ zjWl9lxZo?hBe@Woach@rP;jX{zyA@q#gWaN-DEIGlDd@Qed3*_((+<|t$n!3bY6Ke zNVz&fh_v7B_Ro74^JXV|r*Wo?ZSHy#^R2Y1r+*+ubO4Y>zN2(JnYUu7c^pTz(PqoI zNUW;idqJS3o~K5i5V=S2TTgZW$z7%-{BO!#8}8YXl#Kv?>GHD>3sZ^UYkemK9TPIS zN=eKMfh2IjaOT*P?O!9Q!+`ZAp-2k9GmU2rWI^2Tz~Eu5)@o`W)UWRK+*@X@!Pb)a zoPW{eXtrnx2BXizreLHpt;5H5U-W13crdTAi0pN*RY1oE1E%3@rjOxWbOjzIcTEi7|%~Mzt6( zEo&ob0%TY8sgAK0GCzz3al@d4Y&@`>SAXSL5|I%!AFy(){s=jz8MCYC%v7hvy9x|6 zm%1^+b`X?`PRvtLrW&ys1F&;X;uTaxVux#?1BQ_V=m8B%rb5&NxYpR@)a+S}%PqVf zf;?5e(p=i+Avs2Uo`f6YS8RC)!sC`6skK1T4E53QmP)+p97Efv9~6rg3s#e@s(%#M z;POJzY?X<73b`f_@MHBDD~Y5;0eK24 z$Vdy-b_ErYSAkvTc|i~mU~oa1qPi3l?`Qit{CC@W2@`Cm@tCnV$eHgtg9Gsp`ya|HmN0W9TVf{atxGv9d%>J7|)IWZ>sV1 z{b<}D)cDSe>+lvGpU6Rwm{*+`@Viv=cQ_%;g{y>~O1S|LXEAp%Fwo10LnCG=8RyVc z&4x@_Nbty#a+vsG6wPZ~Ab)!tEAaTb2C{%?UD?N|A!her75LNDcz+jxo$3EMa4uAK$VYK>eIiPQ`hJg$u$5G>KfG3T{jp|0bGRpwX5c~vl zA$AiBZq)wg!)gogg=%e!qZrPfH4wp^Q=D!JMPsW6Jb&Rbk9OT7wMb26f#_ls&y!}g zOLxHpm#Wmd-upkq-o3=Ad|g(|OobG8?e|}P6&zW`APG8{#T@mI%oO9Or4gU~%~(EV zBN;I|!6>i?ePM#7S_w+z%jQ~rBrQ}!^1g@2+vG7}yJ5V1F8*2DO?t+>DEUZ>5=BX5|kvoBaNTx$;b*SSfVLuN_?YiwM8Yfo%DyJi(WEM z=JRMUO8n6<9*&Z6H=gxI{wR4$$SoAc{_EwN>sQ}jzkTIh{qXJ|@2|i6%LluBwn2XL zdw;P}&Yw=*;>DGmwfEb9yneUux5Yn0gI!-QqmFk80ATPINa9#oEAG6su*2W0s*P{1 zuU^0X@pZE!eJ|vw7u>@MOQVQxTo~@hcd!25ev?MYGEdsq%KZLE@AA8MZ`yvR>s^ZV z(qC}}*yw9OTYLF})<|#iGY{yEwH@Mdi(U>A7}oiT*W$WUQ;4yIEiWT;(>BTiobV3$ zn)nCV(4t?s_o~elh_)J0Hn4i(YmQ{K)9ZG6ev7^XdTK3aMD)ea1M}~x7%y9kr{~YU zKKn0FO9KQH00008084p)m%DLuMF&ns%xO+W%(n+}b8w*uOL>1>vT{t{mwyvf8eUBEz% z2FH)x{);$&f%5f3y}GsQZL`(m0q)yaU zP!7i z?pNH^qxX2>Rryu7b60qay8^_3m%4iO;m!Hmn@UW!f@u1M2YD66OScmgp$0l0MKbX# zt_)lbBY=;8$mdFmhf6`h-vA@%vb- zJ1@QA2clbfT2g*-Rq(Xn;U+7eCxfTtdT4lRgvVy{#fM}$`Mr}wpw>Z(*F*Du%x^{ z**c4VHu&g-GC~?t)Lf{EW7gkoM?+jGUdQuMI>aDjo!_^oJa(c8jO&nM3(ZA?yBJeq#Ll`A8SJn)bMcsX!yw8?9Dc$q~oR26<+RkBD;Su z7@qW%+(u+K2&`zlV7nzCIzHY%IC{2~b`3UvAFzDh#ooH!m!w?w3k&iB2091uqA=tN zoyPKm1~|fX!IHJLR2e2(d2y1^E@ZbFwOVNO`N8`7g8!%D_V4M~0#nFoiWpJhjC5-< zVicAyVCCYJyEEa+@legNEXv)JYUhP4IiSR004s!CR~;hptrp$&ks_dqcY8ZojU{w{ zDC+)-ohY%>@`ei&@EsN>#4JU(uHDnUySy}It4f8{0rvp@<)E;ToBy6zZlnf-;qc^< z7-X}}s%6TSnTxT1a&kD{(pYWGTzu)R{7vb!DJJQgH5@O`C{>jz&BcLH_>v#UIbw-R=L5dk=ELlG6-d4BkT>K#{YO6dlXy`Sw&L>k5J zyxNM%Qvq$v-+9lE00Dek=P<;8{MDgRCb z=HJUCOvZU%6IPg!|KN<7jilBm=i3ft+-d!@JAcr?@H;9prL(P!tnS^x9Vh){aZ?$a@P%it-o4wu;|sD2 zZOTWEJkdIyFCpedf@#2iq;&PxrYHlaR{2+}Ve^mIE7f#`QgEGJH(nC2@G2jA_FjHh zt0!aa8WqKOHhV^;$ruQf#;2PL_HO?o>iBwwmtCmxFqNY$kT%F)!M_0Xm2uyMfDpvq zCqOoTMn&HjAq@%2yp&bLC{D$exjJE&rS_KJ|MEl>JzpWc|5e_V?Y51j z>HB&Lj)9rjh9qmV79$xTm8IH1x~h;=O?AUDEK($8R%B5vN_I4j0P`As*SGT;{V4M! zbN=OQ9TKtP}j;5VZOFP(|IoDn<8!h(iz4jkSK_KgGy zCTxEzWCo%6F-T+FN)4eRvI-&e)zl+8&|qs!07B>>^9#>x6q>iM>433+0aF>Ki^#kZ z(Xp5;?$bGcpu3R8Qt4*w;S8lmP=2m(pT`MeKTfnI&V*#%F^i?zsOsq`J|A}mbdODr+5N{a1`bzq(%N*IEeZ4*M(o{ z(^mYcf;pKd$HZt#g^|ym9mEqoin4XCRq5~Iq*O*2y4n`cETV^owPc6A(4%NHjeWnu zd%OEWmy_P`tch%RJoe(XW^69H%xYd^z42(&i$}GB@k~#+>sVIt9F`<=XBT)M@O4cV8%d ztF18hk*o>`M#M_bQt9p>k`Hlbh)J-(dCyFO2*vjdm#EmECh?u@h4`^?A7w2e#VwY9 zaD=&ocQ8v3CZd49!cY^+D1pfH@f=3+(@v)woH%Z@_dq*KDd~okOKt;>vNq%Ab#v{b z?qJlN_A2&p)spV?Rh4tLx?MO554rGH!j$zenTA~P2yWC5Ksu08^T?6S@hy9B7 z@*yww=LpkXr|v#u$Pf1L4WD1=_3XfE)95^ycH_~LM1S=LtQm}ty27#_KH@k-A_`9- z_`!6@)a)Dl^`K`0{;iLK%zQ1%e&X=3HsOIU@m?qBb^P-RXZ@pB5@*ZyN5PuDxcYcx z?#quy*9|;Skc(^$(Q+5nMK$z)+uhNy=2g8%Hu(-pZPyPCsyjqP+n9B{?JS>)Y>Q~O z?f}U~o|nYrx7P2Tyrv&a*uBc$ZW8w0F}-TKXJ(U*>9kR#pN2R>9E>=ytbRGbkOEwzx~K1O`LOu(J}Xk1tv#c#rom{ zPafox)t*789gU{FT42t>Zb;D{l|ag~&$I!`n$A5irz4rXCRZ}=KTsn~StV2o^X6L> zIzvX(8m=ouso%Bb*U|KVrfp9fxiCKd$#J0IjD>L?7;}DC*MBq3UprLz@y?|y< zJd@2x1_)IhGFy3j)MDlRtEBE?aQgBt@aGUnm~fGr@&OTlvlE&pMk<=!3g<4sWG$o) zcwlQRoCf2d-FcE4psX>q);VzmvPuhRm4sjnyk!r(Akj!eeDPQNQ}{zB9!kLr&py8P z!$#q@6WVt*IrjVK{XxI`7})Ud64?8F$}v0%H<6M72SF98I2@oLtTn z$)v7kD7Mo~j7*w%Ncy!P7CJu+6(5u)1JF!Q3^{XD4baS?3HaR0L=;>d=u9%0cPO74 z>Ux01i>xPTo|5y==C+q31k)BSG4vPGd43A3@whX8?M87>E8%6Z8v|Ba;LBKL+@mOW z;K`$tpHf)JkW8d}#b1}W34Sq+@0_laX+n0GAek@-#x4zy zT>vnDMqlQ1suoy9;64`|1W5c$L;yTT6TUen@;8%`-VGwmPER6vH1WSmLepIzk0-MFQL3B3__;9kUe;RZM$slB7FBlT0#JAjwp`7g{B~ z64Bldsa`iCQI_l6@=%Tf8NSXZGP}ScD>J`;fREWgb2+w5$_mety95$$X(~f;*{_lSchdpp4ei9k&6sU`U*{*s{P&l- zMMVa2uit%yS&d<(z+VGVsd1Z2G6|GlmVi4ntTz}Bqkj8I)Bp&7LZ_7DlzY6+ zVT#tP?L~2E#U_;p6y?ONH0Eis3(t#x=bq{BpXEAowZ3+|gWiLVdVL}5MjCW?sDvB* z(6fMN?lz$>aL?Uk`}YI>6m2oA!6RS&!fxow9Pc%B7T#ERh&yu4c9Z+mi5q>up9yLzM1w0{M+Z^E0fmp)0`MmE`)}9^?4?2yG-WY%lE)T z*T?1Got&N?7`qa8MxD{9;u70`OWP}(F})N#slTSdJPBND#e)`lA3cLGkK>O=<&#l+ zi;nz$&<}zNz)sZ`-JRc6R%zPx0{}@gnYuKoOxL#+?gxtmEnn%xD2e%-?<9vgeloHB zt;`-TnxV3ax+(>+EHi0}F0o-07r{rEb8;7k1Kj3V3DI=wv+Kj`=vOL#?{Hups_hTQ z<7r&sGpP}iRkQe~@4m%DkUKnL1@-Z;C&(7@{8V{NN}eY9#)D7^jXJ~ru!?!1>NxF8 zt0@*>cRl&TVrswt!n`#TwK(>Mc^VfO}@cR5Ehd7Pj@ka$(pB!%;L0o))B0 zQ`BPYkd6RAhF15MErhrSV6&l8b-Nu*KJX2!hf?X&Ta#k-UsC$NM#i*ZU3G~krt#+ zWBlg8+@A4N7;F6}F%5`m5lZA$xx+Tkz0=A*VP#+P+~VYa{iwIcMWsf$f7|II#=@6> z^w+E8^MP7@|HG?So7KImle)FKxUcbG+>OVNNGAUt0*6|nneJb53abh%MY*z3aPpi! zO7@-9qfSb;brk~%TyoG*W)u$Z79wa_oWYDQQIGfHf zWO!93L=*sj6(8~P%YlReFS*|2;KI2+qV)@mc^IBwrE5kwVJH`$k(ejwmr?P4{SYDb zU&EUv_f#A#2FGKAV(Dre7%p0-$zpxru-afW7zEwUlc>gHc432*5yv>^-rJSp1Aj0X z&f!%rP2HiRW~cA)W`igT*{Me!e(p9t;ZHvL@}%T{o*?Lj;lQsTC#_OccAA@*wUGF; zg&2ywUaImJ)x9G&;8HTj2@7bGt*LA+7G*X{#yZ?4sKrcdSdobX($GxqSt2uBpwfc4 zjuEw)<}?oirXBbSnvR3s)c2nhL)xu>s9}Bu?PJsn{qxcA5plbB(Hh~bpokk#SduJo zD3?HgVvV#|U|3-q$f?ii>0hJD`#Q3s zhO4;M@!kYTlt~bFby8Y? zOEpD4ORuHe*{y&YanY)j*9=n>u8fV?&?|d=?YJz>TlGp`tJhyARTgTL_Pf(Z*kbQA z>k%^qdZ+rnBUyK5A)J+xV+k(_e(?qL+Z+k{|xr$oER{Yg-|MxP91AI%X4Lu;K1d7WG0sz zKI}3ry5PGj>$t;t$J#rlTi-?)*q^+;Jak^xcsiPP2T=u)+CDMG53-wktjya>3M7`Q z;oQWHZlX#PZ-`}BueUGWibal2%-7U(5O(AKD0)l|2{{N+ft6sg9V*o=LWE2JHHu1z zmQ9uJ!#syXDVp3tiqeLduvdJ4J17)oL+0-m)KF9TQ=2^0+gPe|D7M634!pb0{r)&U zuM!3O?5_7Yk5%a*QuVb;v6iMA^$L{fl0%m4gkMNR^C z*PY%|Xy;OU8A3)ytpHM62w;{aBMJElS^Ig)Rj8uNi|k*a1mF5gFeDNRy#)#zDu`Nk5+7;<;y)Ctu`?IsmmS@9hFc{4I7kCwEi#-JB!#0ItC%M%5eFvxI3zqmn9D4 z&UokRB%7jtuMz{{(guaK!)D#tSt;rFAhg7N8Z8k}wblkPFsdt_xuV#_l9~paKY=t; zhT5u@fa20vn=DeLKcQKn|G{m%U}vw8noy4HUK`-09BWs)N3K$Y`_%|mA2+$49VL#5 z2Alc4adLX<>Hww?y%i_SM|y;L<*t*1I=h!YZN|obPfI>-%JgU2fo4A}=UfWjHXuaU zS~DqcG))E!Qi#)Nm4)9>6c6|hIc^!3!UI=gFuqrGhyTGz9-Bn`f4>zFwrF#mo^dal zu!-U5(@?F~{G2&=A^Q{`^)~)d)!SZSabmO|ksAo^Z~&F>7ApLs=C7OOR;)r(&42za zMi6*^)qv`>Tpr-cBHO5XArX!0dLYC)t0t*#Nr-OT4QrNY?d;GSyq}jdVVoE%`8zSe zUdC%I6pXGBi)0&Zt^fORNib1Wh( zKwO^|vp?lCMvGgxeMt`>W#6>`@1wO4)Atd7LzxHzCz)SR7gl^RJftX89Wg1%59|ir z(C%qMeZc0l?f3hGIBv$>LEP;7VY@l%js0dc?vA^|VW(3GXW8v9{r6?E#7cy^rc=#i zr(UPuJMTTh6Mo4x8hJ)Fr4LZxJXn+x{0Ni%AngxH5yQ0&Ri~Tu4zyylOqmGa$!g5|7lgT9Tvs?WEPC4%= zui50~*7I-uwBBOGLw|VQeN0tnFfJjNM*{Y&aej)#!U)43kGDeNiiMGVUkZ@}%s?2I z600^NI86{y+SfkRA8wU0U2dsvWqaIz8+4*UI1NX`et*~=b;g6yU^?gx2JIlKB7iQ) z@qcSJz5jao?(+2?FTa27z53zDKfk~H?hikSrErRg=h>ZDDCbwFZZ+kV?6vog|Mj}< zZm%i+GcfLTVTZK5mrzD^tlnkN4I`5DcT zUgU2)kT=$Hh>wqGBSH2df8rg5Nw%E(+Z$kufh8|!KfJPYv8y&>kHIrS=eSMfPLV&f z;o#7(ilCO{&Qhv$Cdp@zd@P{<+`Bi8#zDdQVc za9yyu((JWLM9O~Rseep(Ayvl0alaSvvREn&XuI)RW=Su=KAk_`iL3~QpcXt~pCA+j=Ys2u6-&Wp+{oB4HD{%f*@D$nS)th49T99r z7!2cv<%AT_+^tn0#Qv?)H}h28dQ9VpT`#3!nJUCgsT(W{V}EHjG{(Re-|&TCA!k!9 z3K^Fv2u|6p=6NnO<64!(431rHpqY%Na;CZ7^ohZR$b=>aIZHI3d(K)sX^|_h1-MkL_zSgV7D?&M?vr`VrMIRjx)noc`6vtaU-w*UUDGSDi+2tSqzA^ zjLA$E5@i2VgMS)J6fy&Av20`x4uCyNL@w}8R-_xY7Fy0X_8j}I1YEIUg)qPX1pvuP zBn*7kcxM~0`$Vios&Z@j=g_w1G6T!;5Xr%JOOyXsk-RwZ&;Vm*2x!xhO`W~_tjh~m)#EG zBiQLW0Xj>P5DiE8jYmwawrgbjk8yMM$~0F>VKjlhYPS1 z8f*MHA;~!s+CF^(LXr!Gp)Hk-&(SX#{uV=a%{7V+TrVwG*@%7l)O>0E9A_QCG0|r* zKU&Wbduo0GQO5cN)8%ESu&5>u)u7_UF_LiZ{B_Z`uXKsW#+axx_P@BuNM z^4w7I!8glR07nn3KS9Y`$bVAkVnTaxz@^|xNS`|aTQ!tqZ_RHTKX?V-h5JL(Vxf?f zd0Mwi#y8q36T>C?7?B}o(t;~awn~%~*HkE}@3rb_<-#Yd#TeoNT~o5L3E8l*y1gh@ zGoc5y>zyTNPaqByilcAZR3Bh4SO_&q%GFAOS*XRe+ACgY`9A!Eoqr0oK%2kZupxUk zaKU)UoNgA>9_x=Dwl(eUNgIJ)pKw5&wA)XAeCn%GIIIZ|*wEp*?^ul9&zY)OAF!54 z^rg*ZZBaDJITC;i>&lXD39*8zNu4@bD*Sf;DUw7h&xQwfg;yEQey5JJWjB0_kfM|D z>w`XB_g+d3NJIEX?0@_Hrw;#D4b}lWpjF@tnozmNPdjlGScK0A%w2*ayRfB@ckh}* zzy3OKf{%$R46+}6-w!M^6EHKWqQjIk>gcvL>**a{T^JtzCPql^E!p zs`~o;8@CRT3Yc)ca>OiwP85gNbo=)e{srq$w7Ag;?=IqRsBQEk)B(T~K#u$)6-|#$ zKQc)CHgtS6?}DSh7+iO`0x@TwAvjP^O>J*wsrKH_m>DGmqwG>MfP=!CfFqD(*6ug& zvf4;l*G=V&OC@bzj_U9LhgEr-OG_;Vb&19s!xlB?Bpo9xn@n4);$bU+)mp(V3}iOdgl20lH1CA+T_>jZKGXoYwIV72}3+#Nde8l6;=7# zvn&UVL-G9RX6dXL+@%xFA(=CMQsVX^!5)Efr}z>Fr9~CCf+t3oKx!zP&cphMu5{aU zVW-{EpGZ#!jf%-B>!@a9G5d@{`x))f*iPGf>y~ABtI3U=YmvS{K&vToD~otdmS#;n zZ*fJbh_IurMsO4Yg-wHJH97_j2?uTqj9SoG%ftb2YE>8h@t{bI&kQSlygqsEPhOpR;p5Sq&2o6hGCi+8@*Y32NpaL=g93H=dB4`Px zGOZ?|6h@XuqaVRPRT#oQCNjhOC`w#j(DCsD>?r5NEt^+2Hi8Nw0HBfXx*d|U8=`-*l8V(d z`*#WHNH8{;faaoH(7moZC%@V=a~#uVEC7MjRB9JY6*>tsN-wfS?e`a?7v-C?fs5Hp ztk`>Aa8lW|h_5yN!ZGG7><|vY2U^#+qTjeEEL2#%{4g%H99tOw^cz7|Z$b>)uYWM> zg79EVlAf%|0E{MCJEz(tNS?F1Rr*IYp_aGklkasx*Lm5!>avsPTEx3`e{Gn-s9$p*Vkr|#&>X!Vo)*nx85mu<#AcRoh`Uli&BJxmp)tnXOl3VlsJ=RzI=IRg<8Sq=^H*?iT;*d0NY87GT#oc0tH?`|A z@sY^$DF&KWX~>6Bz>gf6A6^gV^qKRxwu+$*vjq3NAo~<1_REje6=X?+#gAo(U86MH zmK~ZH7lxlZFY$iN^Ell8apCwS)J;U8V=8Z{qFw#JHK+PHivjL_Ztr`0XJSRm;9hyo z7-j(*`ZX&@i*te^7t=do-n61)xubJ!_G#fD7Shl^>|pG+L$UIV zvthGEzHTl~3wxL=47ye(E=FDb0O0=_aO`d!*yce40X1R)0df8laO5E+w15)+|00eS z%@2ol4xB%F0PtBmj`5~Lw$07~O^9YGL6Bmr8L2K_mymBV&C_Bi?8=SDxSf3;xoOFC zw4*(uULNqu4S=&~Y800=7UO zUA1!E8QNzw8_;Y=Vxr+xzN_P8(En0rqD8CHYD!MW8iAjTU@Y&@F=^0@5Isz_uDQSD zQ#y}aCNo`;vC=GsVOBW);|EGTfhIaYJD3gRj5X02T%86_u%mCkbp7Kv&G1E@#S=*ysQNc3Gkht%*oI z+}Ni{^t?YkoVQ_e%9_$vsF@2l?P=d?k%(N`$}0va5C2#j zMHRA#@L2N7h@sw*rQ|ea-1Qd~>joVFj%u#=zqTKcB0mY8CMTiJF+u9z7FA`cA)5Ab zP#E5K3@IYbH&52F;uIR6o}SLjT%!F>l`bZQ4T#!Lo1LlC_qnB!eT5GMo&7FRMq6OV z1;A;nih;%~1ajzWqiy1>Uq3I2PW7A3xdacOJf4~@Bwz7MGx*_W9(IxC@JrCPG0VVk z9AVu5LENza&pp<V?q@uiM|?aJ4=4 zBmkfiAjNvU$UD=btUTmCbsFaB0s8tF-`PKc1XbNkif%nx(p7d0Cyu#(a)XN(bhAI0YafLNa*v@Gu>4whSb4 z33aM@x`CZ*PW>9r!Y80lexG!pkuW8kAAnwdRPypE&Cgi5v&B}Pr#Ir|M0c&|6u-XO zWe8vayZFs3l>v%i-7NLlD;g_KNjp{gawKI}&ysA4y2_;b5V5>qk>yB?x zZIoAA-R$3e#-2OW8Y zSU%NhC>Rt>Sh8wi(u1M*+e*X&mWyOa(Xm196zMK$5G-S{SF(L059a}IJ7$OKSu!Rc zUSp5Rsk{(5_Ph#7y;~dC3Kg0c;0A0%IgLuGoUqU+y5TY&Wm(mlsXrgggms|u)QCL` z7tM9-)%J~$sJRk0$tA48+VLH)((_Ww6(|qgnQ9q(l&_Pt6a=pE<;>)VFPG1gR>mrE z*yy-t;&2nltlLu0^iZU_(!zKBLg=ayw`qyGNf!@&uvtS^}X{3dPn6bN7L$pxH zCqx%%zwkmeWB(oDmefbj5 zhQ{}J`qreTa_1TpbPPf~q7BGcMbV8xsL={;tF*neU~i)T8Bk>OOU}7B_vbywW`K&Z z4rAie*@wxSn0-99xIr;K_ST*u=t=|#R4+zxYt+;Dde`QAjl~asgC6ln<2fLjVG%2- zCMo{ybZZ**3@ip_k8rrdBn;iJYO63W;uZ2aF=w62jcOQ#_3B923QT~bAzYm-_^k2= zv5c8u0^7xVx)DBLRv-a9-}Tg2Z1A2C#l?166=*YLBu@OY?zvmUPsT^-a&Yt8!%+Cu zqJWkT7#aVhrLJC&UmR^8N%!vBFR0^@Np5dD$3d0fUIuG+fN^;Q-0`EBs<|~e&q05( zMEEtgY>40Y|3SO*w1)<`g5xKPyF?~`9$W)`OQcs_F1Z}5RfM8u+;G4_)^QrFLsH{? z>=T4J@q1LZR>av8G*>`G#QVI$sReZtt=c(e5APu@<8^$b8B*RKiN|w9GbmPI6{HmSckL31SUPUQ@BhDWE#L?+8ju24( z4otbzbZMrhloS8U>nsc^Xi1E-siJYjx(E| z@jfil>Cx^9xfBH;d!W6wd*sN)S=x(hgCQmwW5zkBkMov2cAa#13X=}*koazC)wD>G znbWsOW;sM8N7s5)BAEm3?X`ppP>JE?(-iD)*N;jZ)|jbDn(f7Q*(SuXY}fj^bzp~3 z6O8L{{+xi?>S(E}2;c9=kH?EJT0_`nU(Z`(^$hF{kc@hs<5D{TDa%q6#%Cr`|e*QeT`Z z=w~7(-7wZDSlk9F0yeeN!@7QL%@dD_(1TFRU=qEU$d%bTt?mO~6D#1`6ASX^cEyHQ z!-9369db3miXyLd?0q8ZvTMG#kEekiHY0tOYHMLF@gZ==b*#3fRx}2u=IS+YDdP7Z zT5WX4K;w)>|Hp^XW_2-$*$NZIWAu>=D~ZTu#1+WnnaTlG8hc_7;y?ZrA!7mUD}ke* zljjhz_9lvgbt8I{S9iaZdKiCI`8~{dq3wwu<_hruu5i-5t+;Hr_0^hp&X_J+uzhdyjT_9EY_zj~lFMqoAC2)k=a z!o<|jXUx>3%NOH@B~6KKoww`_U8Y^NOf6rf5>vue!vm=@p@Tt?Op30WTm-<+nvs3I z`B^+VnK6VAG+ER!%!b7cetW#wYc9EpkWR=%))w9Ve=@03MO~=>O_pL3btwS-TPV4q zCl@#EO2P&4AMLAJG7j>83MGFnGXMJ{PZKtbq=NhJ1?hjQD*ruAfTWAoKW7Nrm2_#_ zm2_Ql{VTQix*Gf=wGs5WGQIz$mAc#%|D~^sKo&v%(tf>e>i^JT>%DGYaQ}kG7d+e6 zd%d8p{+b=tc-X4Jrfs zYpO#O^zZf4uIsqb+pwS!nd$y&uyR4N|C~K6mmTp7NNu_7$Yoi7WvJ5-?SEyf(~vN$ z)6g%3eIdNf>NH~1?XMZDVWiGKW^E~kkr%{&>EOUfzJKWd>6wGT$bY9lOgbp?3sGN) z{zA+bV!sgACLI({9{<!Zrs^A|{K!CRVM< z#{UdIO?YnywGBUy2%F@u2E3YV@sAefu$ugZlrN-yA?*w4U&v^4Sk1&>{%al%ll89? z!**b@zL5QeoVFd9+@CRjO*+zZ|DEdpcnPhIAtTqH?63CqJkb1ewf{T{f633DT;~5G z(b8(=b|L;Ms=c{V|0r$6y}41Se`m2Qsvdp?6bPsl5eSGhjrkxSGi`=G4>C=EAs7B% zbLWNJoB#a~n*Yp|{FiR|nac+H_m&OuLvK#IS(oPv40|Iq*QhDa66`*(ucet(%I{2j821jf}pbY2Fm57*X4EO2)uCpY;o-diq;EA}q4ptl> zE-Sqn=wE)n6{qugGDHA&X#hgnRf`>$*7=ZHf>Z75J835Fz!GGjsoRZN^>W$%?b2VD z@MAz5TFX*X6?At9BSusT-r*H=D>Iz$!gJKVXB&4r8-cTdHmf1W{+2SbO4+@+&$(!V zqv~2eeStKH&h$&TDK=bra@=w0w?PS)e0HCRfXpjBp+M{g+WEn0qb{M|j~ zBJUp5MVj$=N8ZPFO-V?bs!i0R0*Z9lx^GUEMF{Q zq>k}Yb*vvqtT==~;GOSWznk)Es*Zm^ZaKUNZf~u2tmx1Z1q_!^ml7O2S3NCd7kAsJ zRhb*QVdCvU@*%dEgSj86JXy#f_TXhvNITNHJM?S3 zaONy_cqI@ll95hDKX#V4@ZCA?Kg`&x-{1H6+4lZ;bOb29BgTsh& zizEcOXTD3RmZ{ zgs_#!8yIvZoXVnquR6NvxWHc^@0T*OSFjJ6NJJi8o0$Qn>eR6U@{7i)(Y^u!jSnBK z#b5^lISBwmuQ+lS+x2>NjBbz&RvD>LTyhOHk7Ehz_n=WLva4{(xkO*_E)$51Z^rchrb*y-D$5Q|38bL0CUlz?BlvC64bW&{$C(#XJ z3*aX@H6urI84Lu$Q^93UKMgja87W-3n(2r5qdPL+ag3#4vtbslR<5cjpij`f?(xSy zsCK9;hHxtd!{=7X*ug?m!6MN(r6j_*4;tVYNxcj^;YyPlOEA6J$a?|@xvCQWjU>e; zT!2g{fOTTi*t5PFQI=yUnlzn9$peO}YCt?b?t4oyH_~LF2sf9uM58i|5>4j;TqyDo zjchgz6*(AjAk5V^Eq6WFN{xZV+u={g=0B62l_>zvWk?6Qz+-wM7wp~II|d$5I(C2r zhFbo2q!~Cxx|=TAqNt63{6|oNT;iQjjH$P%C3xeX#S<%zj3l(!as`K&D*;Lmhn+GJ>_h^@pC`V3%y|MJ;v(rX zK{~k;@BD^_TiM`3mXjOANtXBnE2byMlcOLLaep-2todUr5%uZi({<9sw^8GOqjE+& zMpHu2mXvoVoed8#PJ40d0w`!BO|g=(xibk+JEfh;*E8;OfPz8Z+o(z)_qn_u=4+xf zxP|Q=s)c6}7shC48KUZ$Kpz35nxHPwlNCUCC(l#CV{Rmr#lo~AKH5~SjU47tk&e%@ z`dGnmkf3lh+WkST>eQ;9zhdG@Ql67YMHYYz#3SF$GakF8p$o)Ou}c&?6CB9AWtw?M zuaeRDKoLEQJ*D%U=%pNXY7bRZcP!04lgBKD&B9%JLAvz6vwdjbFO~pgOE>Z=g@g*q z-tR2gRpY?hKskrO{mIM&@ zsBzRKpVhRVucX0KaHx^ua>r35)qv}iGTqb=_(J~LumR_PhXHIdC`pJUH`d0JdUzMt zH=RUH(Ua`UjkM&SM5F@VLqep41>!3K#hJ2;RJ&xbh0KN6P#uWE?H`i&$3g_{~sJMyikNxb1D&-^bSDSO+C-Bhoc%P$rE$V){D8x?`<=yG;Y9iIT6AaGn4QYrExg%B-*fL zA<#yDZ=pDl5Yw{B;v|?PqpdboO zVCWhKfgK670rinxfbZNL2(D~TfxCSslflgX&Xz5OC(I>0AZl*fJhjI$k>nCcxQNR} z(Y%4fhlgRjETimL#AmjwhWE-dh+MU8e2ezeEdbYhIJRC{yfb)a=h7CGP7=>Cxlh?y zKqTBhU=a zgp2PWBp;a+Ng1etp~fjH?pn8=*`jU6)9*-~5RL~BDXgJO&e1eA-7gcN86Z*I$V+u4 zy!`~JmR~zjfNDx4rm(muakP$|+KuN7 zUD*K1hsOK@vt+rk6A#G#sj1)`b-IN{`+7 zLZ{WBppVKI*p>UHu)12EgAvM`T>9VLt=RFd|zugt6=TZ_| zUqDm*40g)0wHx$@rE(5+2z>!J@7>amqc{UxKDB)Q zH;Xw0Qg?BU-{On&NE)^M8lrL7c7~-L z)9!ZeR%7Yt4QSJRp>bGGsrxu%(6G;<;qwxR4!deHRFDruN2DXbT>cS52Fs-t%Inwn)#uw3Hq-gu~^)F}qgVQ;=j_160` z5@o=3t66fAs+Fnlu9(V9Yg;<#ZFY#c5-UUC#?Ej(ZO|@eS9}^uMj$d^wsN;$t=(eH zEvG~}HN@ar)*K;lw@e6k7~npFJ`s{;QhH&rX~!x}2+EzaEoLtCqKX9LJ4WSe~LT0c+L4ili_&WAtuPr2k_c1`1Nd(Mf;dwb{c`dMV1 z*09`_d*518%dJ6OCJE)VX!A$bbxHOp2?34`eGO;u#>)#ANV1p`6D`S2g@NIOAdk8= zH3}Tq1Rtq_W?ql~wG|isVN8PO^rsZ}L9wbX{V51&WrPrNS|@-vKlUeWN>b4^c$px= zj6qKC!BHyv`{*Zsw{)GHV6nArFd`i2x|~ z?a)}3RY&=sf#+TeA+zi#AcwW7;dDCS2aAKd<;DzikbckhgV=ScKw3QMgWv)1GIsqb z(a%d4&(e%Ueq;B@P6ZQZOVz<@Zgixb}`Sb6Mm!W)UikWY(1w&vvI|_x$3HPe$THeNY9Be&m!`( zIf(t;dpPMK>%hUidCk5rwW((~G*b_*jYn_W?I+zQT}|Ec#@6Q3$ah(-m}Bje0;rW= zLo;bu^t6Q=p&;L8JQN!Y--4#NL^P~z5>kf&3E0Yyue~otLXl_aYDA_T(Y8&A;2b;d zhwG7?i^ny(c35ecwi+6s%?B7!$P@ccB(=S0vsgzs+s4$dna|liRg5F7kun@&zte6e z%F8ghDkzaVd<15%g{q6-rFY>turDWKyhUJw6A3V2I1JC(SLc2$%w_u8RTz>`d|H zT{LToUS`9^5~NGIUCrRt@X*@@;_6U8p2cH==ZjuX^G{U_<$TQ=A zwnEjO*R5O$aC*=e{QQ_m7WknG@YUaczrNiJjnc$=mgfYC+D>!ZPi51Ab${==Aiq|Z z1k~ttHMh_>8ILprv|7f0*wBGgTn2aMceh)HjLs9RK9J4-KoIT-g7!$uP{R6t-WC<0Lvi0?k=bNMn>8oT;8Y2b;Ci8YC!V z^l1%hooo@GNUF}vyfV#q73_E3-zo!w#tP30$NyD|$>MjV0$1zo$dS~FtOXYTw~C<+ z^uu%gM;@RS-tQWQxTh=402sX13$LHp{zRnkHIhbq&eP(s2q2wkJ(e~{3$3hz z;YI)CbD@sk!E0+U2h_TFz>4dulzl5hC|qO=ZLJ8-VV=#w>d3zn0}Cus0=8(UYNLL1 zrUg;2T1fd*!zE5`&UR;xNBpqDjG|f%s|}c^lqWTx*lT2<-v8RBPFnKnBqc_yr#PF! zGwoQ}8XNUfNf8<0PDX*I&M>vI|4m@Ibopiu`~fpXKEtQp1n|N)k;>Zjc1Sj8z|~t) z8n>wqYGV}Z%PietcRs;@Um;~JHZ=g*5zR%I-GH)I9f(Id;Xqq7 zLC%nK$Dj5hTcUwMba|Z?#PX>}!DiWCp!V6W1@(+3oKFI`6UPn%hVY_uh9>G;9;~d3 zN;}b_KUWR@)s2gqJ~K4}s?^?bx^ykR^f6Q(`B!D=xV=nxOi<3Wrq#snP?PHAJ*1+# z9JCsA&#v_2J`@mx)Pn6>kk=25fO#33LmYz`6$aY`&L6?}r1NB}wU5S8=Pe86L$f*7 z9i=ms!7z0wZ#3HPV9AC@8^$QwiktR5zf&CqhBq5gd;ks8)=5(7_GA>CMxFY_qlhxY zt~NLasu%24*?U_B_JjZ$;_8BCN3=CH1B6nCw3VjHFpKEr>qwJj->2?>77 zxXL#BL;x;}%0^G|?}a2tU-+ka0#5hy%ZI~xd)1j}z+LMUVOP(jK)Fyww$~rMpgzoL zbn-7#_@C5jRDDeFSW}kHRv%(t4FrfE7q8d4!K$zbeB)f;%Y>R&FNw0a;pVedC%n^J%=_2}~n zKcVB!lo5T9b^h>jb3@nlr&{YCYxzcxH7Q>rC+9$9#*Wu=5bTzs@OR1alQcPSPD7+; zMkk;;iL92x4ZG>eatQ`v`?5Un>!q}OM0VEEt@%Mc?bWT^E+H(|fLM zm$8D1{hqf(pv_pn7AsGx#xQM;2?r-a9L6Tf55yp*T+do**>$MNf;Z&w=MT!5?{5!8 z5?7U1IajV$QZ|Qvo9Ep&oVj&j*jwXdlOFqe@(V*WSEjMzGg`PGlgtZhJWIKbTd{z# zjFYqBqxDd=_JG!NN4o+mGX^866sA^nUvRt$zMzy))g;_zt}HoGjXxI`A^RuJR(g|u zf#@l=sqD+o1SjL-rhbh^!yGX)`J5*(Ddvn}13A8m z^HlJIL);&RGPT-j$XAG%7iJpL+SLLKwLpL|Mn-<)Duc%G=v{A(?>gK|6^{fVMKzdc z#DOsZGR-%l{qQK17r^Mft?=NFI*HKLu5TMdI$h(|n$u3ZTft8;ewb=tZY;pENJ6-5 zf~jk~i+cF)qX7nrIgL1MrLLgPH;HZT4ef!(ASd9FHspH5onY3e_%j%V$!l+`iF&l>ce66E!DbG)i@;fWB)^0vL|8L0%w>zCQV*}hbi zfb(Zm)NB30cbFS%3<%SLM1ExpULYm!oG%U-SE5lPBPo2D54Lu8{NAi?W=*Xe+bWS0 zqe^bI?a9dcr+rT0JE6OnQLlnSW4Yo+4l$m^?tNtFZ=iP(F@HGIhv0)?)T3}IBRK>*&3K%`GJGr6OtYc^IRc>Zn4 z5iX2NflGbLXnb-Ehq2E5^sz&b^Q%#63FH0JfxAW=B)4~T_h@qB5Jd-f;4nuY5H@)2 zv*n1d1ibeQq|ieXlw+ThiQplX1lQTLT)E1l`l+K{w1O;qn4&S3HC7ldOFwo zuZ1gh%LnGK?CI#9QrGtYDN6;hA*jgm@B|1vFa;&UakTI{?$kFY&TwF0XqqCd>k4w@ zjLR-wEI!#^3jL@6Rq+cKoYVfNP8=^OU;ABK=X;UCSFK`k1oH;V!4mtCNMABFDx9hy zA_GrrVuwx-gvjNMG7Hh^2G;6e1T>+r{!*Py0UbwQF)|c1C-^)KvSt|m_@!7v zP`uq_%;i^oHBeN<_tbUb26UC}HxKy%>UDVMZBt)JvHJmSS@FwSxaA~*`No^R$^yje z!xa7%@UNm+%FV6tx8y<@kgG) zfov;q6+!rbCgFN=JJqT9{n$m9_I1Rt+K~F!g*ZH3Tc|1%@TwVQzlUHMRMv<1yMXm! za3dfY=6u8->cFdhlw4^H5idYU8UEm}aQ^iQiTW=$VfZu=$h34ssTJ`CNureJ6>K$RKk@Vc$Yc_Lp_7H$>2qeRl1{`w&4IjrS%du95 zN86-nsps-GylLCqgqPNLb60$Ts^)%7vU3AlIG41WHixV&xmEn^>!kqH(=Sdwu}cBU zBsN!oMsCVE_w^4(pr!S!GjJJpp#Lm(vAAYZNkoso?*4)>o3`WvUID^Wf&!;$DT^I3m* zb*wfchSz!auS)-SAO?R&?IAbaG+zV*U;0WQv zYNbrZw{kxGIu4B4N*J#?Bj%H8DIHTBU)!|pYcm3HC@GNs0hZ#pTn>MoDKhAwuWqh?NLA5Fwj+2qQR~%pZWN^+a zTT5syUD9#Im^;On*^gcimvdaZ?wH!mNne3rnGCJS_JcUn(&)vL;Soq&OGk_-2*~hR z6VpC$a@3?=s2v!L7V^8cP&}~`gYoVGq{YeB(yr%9@pCVgYf5$BG}NWkcI?lRHF%Iw zcWA^)T5iStKq+GryCKWLLCeGMk!A;-mzR)GMMq?<8_!urs*$=Zq%u+)NB-l{q*NSnA(L2W%EF3Hw=7aRo zuJy}M;w=!NGc|A@drN~_(45-Odg;2-GJ00?BW#E38#n;yd~aDP>XxS;Qu2UAQw#GB zkzIyrk>mFqe~lBjlrv%EL1T1ZDZu0VnP+Zf81uvAe944ua5~!5GLhXA%Td)&D9c?aQz!PvY z1y(5D83S4`)EwV*3P&Ccz-NPsEi@kG;l;wF>VFCiRDeYgo=)C}5=*kE9kCSwLnfX0 z-pq&d@Y3i$11h-K^?52OO|hHVmDtQz~3YgvH2YOiap48avzA?}ZS z(#v2@_EH4%jYHPYr?mMnN5nCS4dp?MR+Kr@_ZvQYh)e1sA?O1#KtWZFB|z+@l0&Ab zzl@rRfZ0iU2Rp?`jkCru^h*-0!=qjf(2dQ*hiK}j z6?fmu#aMOJOc<3=fJ}+H4r!aazAA%8)xb(>bm{skPZ0EL5^l=f??K_zqmF$WvH{`F zVc!+kj>W7K?#8~&$>)%gHoTRFyp=Y#yHxB(D5vbq!hO%xMO^Y9p2=hjnimIb zfQ)F8P`mltY3ZlxY!5{M0-OBX{DJO=7{@^3DeNO{WT=P#%Hg9onU z501M|%CfojT=+4VZE;roT zzll{G0S?eP0KVaJD3~G62Vo4{C}wE#J{WVA?-L)~#t?BL!+8%vyh?#FgeU#@9H2Jl zgabE(d90gO`{FMv@{Yku=|?q*T>Tg)aLEY6&*SBcr|c+Z!un#IeL@2#*CUco;ekaG z^&03sqWvydJK0M4;eRdwGo{}zET5G%VumoILz2HM13bEvmqb@iHjyuaS_w2scA0sk^eD8V&tp!RzV z7w^n$2so1{ANtoPcol3bQ|J|t$@AFK@5tBZeaPojLO!_G$Dmeunlq9{d&sVmKH4a{ zj}Y|8U5fBzqFtYZ$+0pk5%XWPe!=FTr{NAb41hZ3qqLy<;on9R#bbCB{c?~&?S>Ag z)vZk9G44_jA|aEJ<~%HNWs5gELaIyz(-G2|b}>tMWsL%IZrO)UT9opHq@(fl_Eoq% z-Us|O^b=rsv-T`e*ibOXqK}jOs!Ykeh2r^4IN=#Nhv6J}zzk3f?x96uVSk9Y%yIVe z^#QK+lX6kdSy%<^_-_8@9$1`w=L2J9?)SoE0^PQhiwIC_S3v-0M)Y@AY6Fk0=s^Lz zyWK;}yVJTEnKbV!!1_?YJ0HOxNQY$o@U73E41jBeHztVD>0K?AxowQ_U7Z<~^~(>t zbn|(-?G$ml{d{z2BkysoVCGU=G!81%?w<|oEQ5!6T7irJp1qQbs)H)aRx*{L*3qbOr zh4?&{IbK=L@jO=WLk;ht3Cm8EgP~7na0daUJVKl{g?}}!yjRx$v*uOldnPGTWIs0|66{}}LMrDv`$)oT(3DQuLF_7~C8>)51 zNUeeLP6uJqs6F*=g?y~~Nw5f0L28toMmKGwWYwc+C$*$#7eOlwqjjW@oBl`K%%+^) z+)(;4fbIvW^vRuarp~P7=5|T zyt`BqfA9p+!*aGfDr+u`1q+7=RhzhnRSk_JHS(E)e(p)# z<+?|RBU*m2>S7G-l0^qoHg##ts|3#oN%g z=sR_~!&3UDTLV|G1=wBee?AK86#jh0tgxK2u9op0Tt4YJ zEZ<^ZpzXvxSmRj*$sbf{n`9hre+>K<14H`Jw*-AH2x4Dsl6`Lj(o=n_c}D+(rw_47 z!%5r6tV(o{2PbKzy^~zCkl)W1Xh3+6O}10h!a0?3T_@xq-D8bueteGS;NlJ{u4tD5 z9Ol#o{fDl-ba@uQX92`1DB9xo>5Uz!u+C!|o!Re$wg-}VJ0YWY>!T(F4lo@4WE_f} zPzH@jD2&>#;lzOseZv}Cvelw>rUV!c+q>fUK+l#H3s^hn8(?0r!n89sFLHfT$5lqqv1hKezjT!X zzhsmR)Au)kz=zQf&>0qm`oE0br7$gRw$ z6*k9~@H=mpu0oc92OWCuQ9t%Pp;Y-5BBJZ}iwkMOgYi!|O$AAs3P9p0e@MF+Nkj(O zG9U0`8QY$Tw7Jr@D;J!8wRm=q+%{C?Gsk&O`%PckwGEwn?sdS+TQ<%58-OM>02L1;1Nt;5`~I>>9l6$u9gwVm zHH6F*o0F3_oog77T!GFf1KD&WkJ-sa?#{9EotjX>zA#va2O4Nny-R6KlLj$qKGd;P z3pr*0IJLDT+r=x~#!NhWOR9hBrCKloCO0Jjz<3{fHm0)YKBTuYxQmx_V?jZ12;}DW zLZhO5P~jH&QkWce1J^phU1vvm>{=Q2UBGtIG5fyrw|85!-c1u#8B7~q(zgY2d&1a= zP`isiQp2F-%}D5*Aw?18=xJ!2#fskOiz#!P=CZwNupi3Or4)#%=xh+h5oHUZ}Jxi3r?H!f_b`H;}zzbek z)mMzCM!4^6_!gSflhvb78-V_~t|BF2!1CNJNIQYPE+~+7Gq-*10SOs^eIJbej zN0NcahdW4Y9GXJbEwo0M)f&m!*O049_W z;2uiRHp$~o`0!NdZk(+_NdhRy?Q(+>;V$(EgO=FO-TgCOm7ms_Rr>VW$v*V(<~%Fc zQ#!gfXIy$WMS7Kh>dX-9Ozm>`C0h zR>$3ZXFpFGze1chF?iiz`--}?=SBq8UfV+aQFGPNKeyNIw1GXloqF^n^|klSZ{Jsi z+nOZ|i0x#v-1g-%+ald7yYF0Xtk>uFV~cdL2j`i7{#@gE)R>SFCZ`|m`Tq6u+XTDj z+|xV%e(?~FubZ8F)|9-EiVvP^B-XVD_!fmcrogU=e_Qd*mivow$pKms$ zmga4~UE$p5s5w8retqF=#@OapD_>tz^I+~XbB)Ex=-@+b_N0B=7Gq^LWaW!KuBT_>iH}d*^3K@4erde9!IpdTSG1{<;$CgAcNNcg zKksFE(bnTUUY!XwPE6god++i~+|ArRKCROqTlXEZ_~M--S9)v@osi#rbfG9zj^*Dc zVQ--cFWHj)vxUarj@?-Cs+A_xkRIouS?I#jj{`N*?M$6UX)K4Zbi1T>Dz*20jj6CA zL*%Rf#~oDr_h}{=+I7g(wA0^{G%80k)R4}}(Ii)4?e@#lNVnST&C}Fk_dl(g>aJ23 z+|$^V)R5o#;dozTxH~C($$d@FHmt2HD(&!+3fV1ew3AF(I@MV_*^sX0qHXKW(mwuL z&yuvfd9A-zx;2Rp(4N=d3v=?=*?m@a95-PR$92%(SmGXWqMDiy(>jzil-(pu>#m>Q z(Mejvon+a^lC<4xv2H60TEk65*&%}VNNtvGoTcql(ujOCNm<(9k}U}zv$PbOD1M>X zoc%Ft%KWjbb~+4vL)=yC`;3?4j68v5z8?B8y@_ zg-CILBAenMMGnOwio+B~D2`GbqsXN=o^4Y2tI>Qm4AY1chP_S*jX04V8j&|ke^1+k zinA2IQ=FqXPjP|bBE=<&KPdjBxJ+?{BA?2|4k%Deh$F#NX|( zfED>@QZT5bNVZ?)f|Yey`k_OCuOaQ#pdePCKH^oXv}K2aL{C;)Grd4st0prFCRnn& zl+I1fpOe>)<9@Ge#I= zSPI$6^>jyDvbqEN>$;WSS$2Qj_>wKz4SaRWsZ;iq8~ZD85pBqxeoyv~*jn(FzJ+40uoh zR0NfP38)OJfU3Y0R0C$9I;a86K}}E#)CLy75?BFiPzTfn^?(Z4fcn4|*a3Us02%;C z&=52NPQV$sfX2WTGyzRPGvEfA19#8@v;?hy2WSmEK^xE(v;*yd7w7;w0&ma>bOv2O zSI`Y~2R%Sf-~)Pr-k=ZY3;Kcnz!wYv1A!kH1O|g4U?>;{h68^P07ih3AP|fKqrneg z3>XW7z&J1-1cM1+BA5gwgAfo3rhqUo6@-HbFbzxxGe9Jm38KI(FdIaJIbbgM5ySwt z>gM66`CtK92;#s`U=fH12_O+H21~$FkOY>24`hNYupfxv0LTUhK@K;Y@BpoWCujrOf_9)i@B$q`N8k-QfzF@{=nA@l?w|+g34A~=&>QpteL+9aANYa+ zU?A`VgTP=g1Plekz;NIX0>B6`5(I)#U^Msvi~(ao5Euu>gJ3WLOazm_WDo*E!4wb% zrh;$~0j7cJU}_JK^01@;3G901wiAjkoSz+rF%90kWfE;tT;1t&lr_zj!{r@(1& z2Al=IgLB|KxBxDKOW+UiC%6o*fPC;5xC;IT*T8jf1Kb3+z-@2`+yw z_z}baw(92Lr}-AegT`o7LX1yz*evgYzI5QPOuB?27AC>un%N{EIyYmf4mL* zKrfsr+Rqz46|3nZspC;!kw3*gF$&gBB5(Mtn;T8Pxmw_(q>$cT7hS?il(biD?8NnlvB=$^K_99SJXX7xkyFMcShOixJWtuE=j{59$9uJ z?512tMSEs5MSd{fU$%-CQf{OoS46oXirma7ksqpP?-}KM6}iA!BJZcjT`DKnb#~cu zVmUdNXp#3r<`(Kjh!#@(-r03QEt2< zH+O-^2P<;M3(MC1nR3e0$1YCfCoAf1r=0RS(()&f4^osnKsn`DIxH%CE?h9=q&d)^ zqS5iPQt~D|EZG}W-qelxVd9E&ab$C`braS^MVvvUG-hJ=?yBCuFfuh z1D2p0ZO_u(2m&RQ7yTEkV5=xuLGPvXm}6-N*??ZoY8cb89>;x|%yE5X13EibR{4^= zG8>LL?Pi=}CPd09uh8zxM1G;vF{_(TxoAZ$e7VRkk#f~BV;3(^9OUK7ap!jmPAkfu zNw@;4jOGZ3Xq$snm38rraw?+>Y@3bHn!a1cm6ebV+XBj|i~@v>^zZsoZFcAz%&Xxh z=xBqL(Nz%FD??|0YAo#Srga-64Z zj+@qwl)m%EQFgppP5JlZ=$EU0xny=LRnx|leOpfH<-)4VpLSikdW(U@)%DN5E?b}N zTR*&F@?Azms?EDcT9pisO&@DtfA;9LkeyAIqa)QRy&^58b3mGOc9Qes^d+3664INV zkX7jSguX=2l788CwrJbju36XWE64R8MbBz8^oz6TQ@&unTI2^RjzIg~8?-sb9QT!; zN!MOhVdbAu>amBTI?6g>J6AO}v;T`16}UbXDsi1nH2)Y}XjppZZwSn8sx*HlCJ#!RIfIJ5aZn| zqsF0G-Xa#;orUl8tCM0!wgtwnZ`zdO^3QRG9ntT9s%M;j|Cg*SITzVMW%IquY|oZu z8;rSEcZ>~Z36|F81^vdw;TSm!s&yi7sp!oo=nwNVWjoQkXpw&Jy`L;{ zD7z=Os4*2E8MQ<|B>VLu--xwc{`z(@w2fDblWUi#b-pRiiq*?j=aBpJ-k2s=;{N&F zM6o&>trH5V8|CVhJ#$ia0J9pWux{AXCre$R*o4^yjJR{nk)CD0mg9V6)p9!PtEu<6 zrpT`h2&I`A0W*!JLpCmF;&~ouV#Z$YYwb-kW~~%&+d1xGOVE3@>f^R=d6w_EHPVH*T(JX!Wf92K_FE)35au9q(X8KJQM`7b&D) z9slDP$*)1LucWDeJ(^}+GF0gbv_e47o%7jtOZI3=#ernIUC4^9DP2rG|5#eRk_M#< z)?^JDobk9+gLu|uRar@P&~{p-#nhv#(1OYw*HU)SGM-7b&D9&OrkTrXv)3d%5mOkT zrPn0BS)o+iL|x}bno3rj9g_Tf_tA}Oa$HnLj%y<8>}fYaN6oa9CX?+kyEGfm9?CXa zsp$cjwPmfcLYhsqv&MgyDpXT_zLQW#R+9BEd#0wW|FB0=7y4&x>R(r-W3-n(?y7hr zSB=`pL*r83;iSg_{+Al<8fg_wWVJMb6nanTEY|8BsWd@e4Mx~=RR zS2j-*4js}oZNqVz=F+Q^{-PdSx~_Y5x5*LFQ-kDheOU{crxzagHKqSAz9GkTk?o0P zORavwSnuk}Hd;H`M6d%nnMS7K?{q9xw7aw~tOs0eO8&8z9g@XUwd#HTS_j3!V6}YS z(9@b~b6nRRG&b3|?&)jwk`0C}Y}w!R$m&?c3FbRQz8f2wbQteho$Bsdi{qT=3A5c~ z6*QaJ(A?CiNm}_S#CH2THKyWyME&@Mjyp@<>q}=!66I8U=vkVXms?yFyT}^(!b)5$ zE+G``ra#3}=VxilxsoOCLSt_#rlYYAvOQ-_)u6g6zJL4)Z1)t#(XX`?`#Q#D=<2K? zmfYEUZrlLbzB*^*u`#j{G?A*YmYeNC%j3RYQ@40{t#8?$53}tVskgA*LonMb@>QfJ z*rEK8yRqr)x*V55?}R$YdTwu9z!m^~i(-Lo*|rZ+V=CUIt-g@(iGEeS+V0NZ^K_~w z$Ax@Ww*B3&xb5d&<+i;KZU2*4+%}G2Ex|NX)&sk^<8OO}0kpx2&rB;Wjlb*BkEHx7 zk9?Uwx>le+k=UR4l(!Px=b`PN?p z8(Hi0Be}a@2eJWwI);AL@2hlycy+T!JL3gc zjQxiFZQRN*7`*%4uv#{lC9yigtC92uq{Tt%p+Z$Fkym%Ar}NxcgBDrYLwb~z3g}Q= zN*L`V^713!0H0TVJ^t*LLm#=1ipn?ej|b|j+6xIeBHvuv&55Yj*&|JBOfL#cDXe9u zUnBbOXH&x?gM-4uLuWMqAuKp@?D(;fW8I>|rj)!=D8Jx!VBd^NE*qC1jzzT z5>OGzA`&HY=jHwM`<;7F-9Jv9s*kE?ui4MD_S$Rp+TAlXOi11YQ}fO^6X^{td;$`j zfBsmt@5Pg{UXG z*e_)IhYT0A^2P{S`M^Gu*buiQW+|`H9^>HXSm5HQr@W(LVodo}h<7O`sH_g3i-Z^l z|G)CUxQ_V<9sjdyN_rli-k)fgUwL?=c;|FiA)e76Ivg?Pi^Y8`9$?{zL5%qmZJfLD zCttMqpxNys;I+9 z{6~r@*l1TU8Z^Wal(@f&W}u*p=aI+wiEsRgjNua?wjMjDENXQ;(R^>aVwZ+J_{zk|aBM;9OqrhZ~Q5I#K%j(0FW`E`4 zVao90VJa-BvABXM9;WH`J+BVDJqM4lVH*&GgQNT()ulhF!)O1G>M*0~v_II+Bj2_3 zW=I*NVq&;#1pRl1{&qWs&4t$T@6M@dk>ROnQCK{~A{vVrEMhUKX>pu!=NbGK?+cHi z#lh*(#leyLCj*`w=I}av`v2UV0U1v~hIF71Q7Ahe0aEZ*iQ{mn4|} zwl~~Ef(Z*|ELboe60AZF=lh+7_7qdGK&{7 zA3D5cBMFP=SR`YSf<-D8X_%IcbXwH8d)pS*-%f`Muw`J8iA5GhfGvAG`rKvXA@Xli zQhrrRUBLuaW%_%N>%__J`&!{{gREB=u%sY(2kzs~7|0Eh4eMD>0cwVl@9==UP57wZALy_{GBU_{A}J{1Vb!=c^P9CA9y_ z35F8k35JrG1VgEO$#YrinUvJI{GSiUl;meprhgV;cArTx|9gOFE2Y%`7w7W>I4LrJ z(VYVa$1ej1M?Xd1PlhX{a#V^u=UHMMKHER1{+~nePiL_QB;wy=9WW~8_&5IQs1zg3 zc{0Q+QhzH9zqulXxw#^}7<(>(elq`VjSUo@;3tDg@RN1#J(uv6<@o<(_2-~rpW6&6 z9?Ehoe}wbHcAl<2d#fb1A~BBU=gT;1DU`}`^x3I(_@pTTPgJiZtH_c6$t#>zMGgyj zEEKR%#6k%R01FTcWh_)Mv?{9nisxCVK2f#(yY%%F)$rF()UZ&;LIVp;EVQuD#zF@R zT`cslxPf{7L_eeOJcJ&ff$^UZm~uXYL*8?`#MHp#4;_Bd%m52REFdh5Fc-~?C%4Y) z(qO^m$rT{LnNB3Y(f#kba%9u751*V0Y>ia^(Ts2%Z4)d^u`t8pCPqiwyjkFUzsmvU zW`Fl{4KNRP4Y-Ac1;#bNvT^j>1)*nU{?{c+&ni4h?=}|JSlD2q^zP{W`=p;Wyz}=M zh08*Bu&~9*LUy{r=TV|Z?4Z9IT*w-@IqTX*#qA@pY2KMWk6gFX(>ShfUiT|b;gTb-G6k*ulg&Nni)_Ubq28Rlyq zy_aSv(~A{cZU%?hQ|2o<;M*OiLJMvK_w4EW{oUMm zWQ^UPavZL7$>@I{yr%I=#+*@=Y-h-$`Ao!eN%_L@9lz$AqHXv1!tRj7I7!-nO?`5Q zm{Dua)OXwTGOC;6-v4Jn^G28{^^@2bQ`Cf{(sY!w+7qI%6xdI1Gfq(xmMYd!&Sp=F za`}h4U?WaZ^C*?H!pM9*X;aGZazW zBSHPI1Nxu&1(d`8W>})QM~3=;54I#qbcvxTeY(UaC~LaJFq9`c70c<%O&l`t_~ON)-D29?G*hI7tWL$B`$=tb2HT3Ig&>m zE8K$$o_LWWH!?%MF(#$3l(?AucX4yv&4n^YL}hmt)p*Ex?5se;kOZEr)r+-|QFn!BI+ zOG`UNsLaI#&2`-sX#bxZ*#6x;~t^I~BNyPJP_g|1M`SE(P__;L=4p-y~Y z7HO;KWkqHKG;+_8{+dQRaOpC_c-Zh&ZV~#Hie5`(b_k6e4X?kZ@eW+RjJP>$SeIMm zs-ia=nca?Ef)z}5;EH90<*;E(Zc(6$-brLO9<3Y=pTDN*4qUa2uo*V&%q@DRqQ?}K z4baLx#|maU-e1;lB?ah{sW#9p*d}G@$EY^Y(^TpIRFz^}@JbrhcUO~QUO@XNZRSf#e9N|znljh~vTYxAkoJsd`?X5#B z70zUM-RG>~vUppS$K05c?909YfTLcj?L7XqG0 zs0i>w!bQL%BvcGMMZ(2EJQ6Aa(vWZo@Dd4?0_YMXTnfBKLS;Y$5-tPYA)#`h2ML!0 z!${~AFolF)0dq*G0$4%96~HDEsss*@a3z4t3snKcyl@pj$qT&(uJOXJ0XAN!8bI*E z)c}eYssW^U;Tk}R7pes`c;Q+=pBJhF%y{8Cz?v7T2b_4}dcczx4PgL3UKj&B;)UJ- zPkG@tKs+zh0HpE44ZurYs1YdPg&Tp_yigO+zza74?|7kRpobT328MZ|7GR1OZUN?a zp;lmp7j6YMd7(DofER89aQUFO05Ko@7NF#V-T~M6;CBEUAJh&Y_~3Q`#Rt6yr1;?X zXh4Y%>HswO;0{2a59$QW_~1^!nh)v%ocQ1_z>^Q^2K@NoZr~9g^Z|Iv2Y&$K`Jf&k zjSubtUh+Y`KnWk*3%uro`hW&LxDR;82lWFzd~iQ7%m)nsQ+)6MFvkZC0xNv*Ah5{? z4FLyy@DPB@4}AoP(fsg7fRY~?2CnhL!vGsUGy)*_;Sm7E4~+s+{O~BC#1FxM20sh~ z`uxxsV8#!R0oMG`IN-z&j{~0k&;;Pe4^IG(_@PPQDL*_3#PdT_KpH~C z@F(CkKQs+A@Wa!0 zYXN8xa1wwQ0Z##F3Gfqumw-nC&@%8;0A2>-1)vomO#og2&@Tm`RiH!wUIktYKx;sQ z0K5jg6M%jIJp%9#U|0ZJ2c`tzbzn{a`U$KEz(0Xa0cZm_5P&xTTtVm;Kr9IV0w@Kc zP2id!ya})gLR$br5Z(e%g3vY~B?xZ=N`lZ1pdkqF0Q!Q^E?_1I?*i6>&>r9SWcOb@Jwe zK`h__3T6Rug&|guSQusnDTN_6@R~5p2C@l5>>xrIW(QHi5CI*}hpqVht30ez7T%eOM%msQ1L)@UBFw6};5{7udr@}A~7%vPVKy;cgi~wH>LrAbh z7)FAxg&|(BK^W!*-w8u}V2?1&2M!BE{NR)@%n!~9LjvH6Ff0IW3PXb6fiNrx;)*~* zAh8H61X7AXDDavHi~`w2AYl+80tiNNBZrwAkg`ia02;3E-85_~EGOM>wtkQA6E0!x7}MIdRgLf#kp`5m*kK6M^Kx6%kk-+!TQnzylFj0mK!B6hUH9SP`TYg_OW+ zqOcOkCJF%{LKFtjAW9ShK`Buf1eHV~Wl%#DRtEJ&Ar;U}6jlMPMIlwtNfcHEJw+ik z&`%Ur10RV(>flpRSRITPg*3o4QCI_fDGF(t&-kaG%y$o8wK~-jKl^gk%=o#2$xxKi zne)z(>y6;@Q=he^HrJE&vM)Mi(@F#Zx%CGrBXe?ca1VY$V>MR{y5E4^!E8XyG&K zH7(n?e#k6%W^lQW%zTJt);}mVrLu3OGaqfZ+nq=Dm{C?*9ZR8+g;N=)W3D3$PZ{OW z(p__2&sbK6RKwX~n3-~0yZbi!a_mnu{L)&OX8TK>!XL%3Fco9z`6|jO8WIRr9!ijU8Wgh6ySMvJquUKtdm`BQ~_at5nK*?5poy8^Tylq^gZrIPS6}6HDD7NNhxe_8dCfhxgg)7#si{07 z-_@Ro$L6|;m;ACCTAePm+x^jI?r|Xb$%`8QY`sMcZ}f{mBwnhC-i?+kX5S3gD z-iL3y_@!<>!PTd$(xPh)9$l28Mswn<5iD1!WNeo&9edv#KeI*X)-38U#ztziH4b#K z)O@p~$|wK9etGq%E!~2&PyhLzTjkN!GxN{#Xc30Xy&AobC!As?SRP;H=aw8Ma1eWE zKBSs^ZO}OkLrR)UJlIN|=@>HV8A7ddmpSe0Zn(8HqgC?bm6+|I?fKC;vuU*Vl2cIt zv(E3pl$Go~zsgpR?b>a}uBEkWIpXO)lcfWPfpTl3#^Ea4KS|w3FN|FOiT)vIa65B+ z_kE{);6XYXIhucZikW)gm9}{QkZ-0Z%#SPZWG};><`?orhKU7QaZs5i_>f2Np^00O znN#tz6z&!e?p6>rYb+J3757_ojrHcXA4>k>oJgD)R$%aK}Y_j#d*7yJoka*WK1}jdEEm^qHoqW z$$F!lWf)o+(9SVL^3M~$A?@DZuN1F&$tO@zqvxa`q_8GTLQBOma%N?}f-Z1Mp}43I zFna4nJmz5|ljiPFN}{-U#C2}&Q`wjXhJ{K-P<{JqqJ8d=EzU-k7b-b}V*Bh-{_;~Z zPPo1?qq{?=7FpX~a!9 zy_HC`F|D`IXH-S|c}~g+lI1l^{>Z!LIe!Il6)f3@|I1tHCp_f{wivOI68z)#*%ks}L^;ob+0Uw*`wZJns-7Zmsf;dy>q zG?I*DUv#8Fm7%ITS0SbSvG`_FWb^j|i+Lo(7}&QXLWlDtMqn&&20`zpW1Y;`wrPaY3mj5zy?12buK zT2a}0`mzw^HLrdh2bw5p;FWZg(;I7W_O7NwAZ{t1euj9$6vpN6M zC34;|hULDcA)}h%@q;gj;D6W{!(&Nw`nT>QtBA>447KQF@wM0--)=P3z$$XfFNL)tMux$y;>jV!D1M90&d@?^OuT4(nIR|n)qLbnH8O7fhncb` z)pMdG@uL<%{%IEGJ$dJg&ldH5#JYJuBrP1&P7=$H#`%SGO&)EA?}*wD%mH$nqs^9< zKOBQOo3uB6=AT)MTS0s*?~l=$?++6kB16bAE6LWLp0g~^r(W=JoeL93JHXb#vsw!Q9JI(`=ySkOCSxz3?u}Dy)V_ZO~Y% znvQkafuJv;#wfWtzjAjXM z|8`p~8W2bS9{SdhqdF~l-7F^POm|=)ihTKfk@8-p{*wYtUne7h#oeQfE7`A{Iju_) zKDT^ae&h4XWH+Rz4_Zzdx}4fJW9B=|c-N^Npsq(2_Y-CK?kqoYpl7pGR#kT0c$r zd{(&=VV>Fq3z?KX(pxxMsAw4DB>rZ<*qGjc(ySj6O6Oi{U2Hm#>T?RZYhbK`jw90+ zkO=!M%~pNudA|3)P(_7=fEO%P?dP6kd)0%0-l-wBjULeTli|S&^NbR0l}4mcz-P_W5dcjk*a51(}!EI1&rD$95VG!t&p* z?-&nSRo72Zj7v}DO=y_@$PcJVtoqdRc~-aGiyQ4X;Y!4iA!|I*|JzE#vfr2b!UIvCO$EJ=giR-BY-7=@T-Ns!e=g{>?{Ni8NDv=oxY9~0+(9H( zJ?{~y>u>UTHsGTcZUW#hJk3ZAHPZH7u@NAlQV^vH^%yTGthZX_`e9owU?*^)Z!yn~ z;eFjHm6WwiqL18mLtp7Km-@J;NUn`{9paBKh<%jwdogyPyp~w?y+?q5NTC4pBg4jPibX8y zjd7)>>7MQEUO6UbUF>1lhu4BSsx@=r=mSKA_p`&%@^ZmP{fl2NP02o$YO9sD$bR>x zxO(l-+OnL*K2cO+zVx)!b?T#gFcLq(giN4~RZmVYeZq_$LQdY`{kt}XB5nn_9 zYIpHvI$NEh88@0LsJ!XW*HoRV-JKU*r&)HD`$w5qHIsFeXLZYk5gzla#$PJhEUlJ? zzxvi)*K|qiq<8cTtTU3%7_6vRoCq@cot~6ZGi--yd^6CMcyNDVgy->Tv8(QWJ=yE3 zl3r#bqxK5qY+q@po6%lH8P%TA6kqu``b(S(rP@Uu#%Pa0wIv4Lq?J6v`a>M2Q{Nmx zUq|yc;bpyz<3X#gylT6UC+nunrQM&kss=ArR?+t1w@h-AR912I3CK-h&JIq{{iRB6 znc8gvT(4p)+pGE3>RsvC&12#Vls>U^J}EHMUFUI^WpQn)sNQe=X>;`xOMl9pH8gMk z{XxrS#KC9ep z=i<*W`jmAi>gvGNz_s_SZM%YHYvH4-wYzLqrf(KrcCEL7tuj{YYrhhl_GNDr2j%`E zKC1g+P&G@GGybKS?Z<@I;_#>^(Lfg3=l$O9)~D@U?IuUMS9g|m2g`NYYp1KFElf)8EAT6;5S)4@cV_@*{j9LgcmNU z`;7IQ2^g;imL45O2@FcVT#Q}!JDHTW5E*(;<=(1BEJ3dI#_u5YPi zE-_njMc*HJoVnR5ohYvHS{fN+gqu2*`b&P!mvp^YWzNO^6Yr--X;-oH#Ai2Vt&*l2 zCA{1R!JEF7d=Vw^0Q1i*zrBM(v}xDCeZ+k5(bClN>e}YeE6Yb!k=VlCe6in6uF=gH zXO(=zLU6x1*FFdB!Srj+7hOUzme~5C^(1HKP4+~x!Dy%6MdUzA*wx=2a!2pEur*Rk zkY`|Pg_uOv5zMSkBAa5KztuoC}=z0;XTggMtfH|NBi~S}m2kx&-1e@Qoc%H&w zVt1Bg|F-Il)2>aZ*mR&pO%{2rQRJFz64&Eew~b8as*o9r%;Hn@^LNKD3scIP5o)_C z3w1Z%s3u=sdseSOY3#P`GgiNwNtRK6(C9QRAKYH*)_Q+ToV=|5z<#X0q9*_T{H9C8 zx-h5f)M9L%>+gr8+f{}c3Z6ycBAeYbONBQS( zt2=>fcFLw{1F?Iis-am^$Hhm&?B)H70Vej_Rl*q_#`g1qJ2QF_$+(oX_S@kkHRg5E zWJ9Jkb0&fx7Oj=UzZ9@GJ*Q<#r}vvo7>IrRaTy-_@V?rWISFldHTv11Lf?%_`g%=< z`Pk`eBjpU~f@gdNMW^hy^)ag%mzb=?uPYjpc~nGoQ92&wIDdGb>>4~1WJZ~6yFD+L zOk=IJclPP%j1v>h_vPvdIva0-X{k56IDr?9`=aP~N4%D=c@e+4nnBaX)GU+Tm8Yfc z5h@>X5G9Ry9D6S4A8KzZJz4A&dIq$Qxh7INwd0J$?&)gB@@=YBTz`06@jNZ(R#a@E ztyI3PUrxfTq1d;Y3&X4*j$i4?+&N(U_G*UEgNZspvHrP&^mH%VbnkC#bE`^|!_wxP zlg;ldY7yb4yc<>Hy?{_J0YZZ5qkndP(ON^au{*DenFBvh5xT)*s@OsJJt)sGnr_6L z@AG~1&aI_GvDnOh#*uQ_t*Z<~HCv9MlZxrjV++ls^3D8eO;Q1tYvs#_`Nwz34ckyi zrn-wECC%i8;?o!-yWHTAPOgRMAx;j$rC~pa%nxcMk8LC!Y18$V{aXK%xCHvFUqYb4 zvY*br8?L#06CJ;yRKqIV;$l~X-yuG(9u`kFw&s*=xss8YMwrZSl23@~88M4u&F($f zgvd4_c|5%Z3cZEA^U3G;hO^i4*z*Ig0*w}z+cPYeC|DO_$>O|-;?vxNBMWc2z*_2kg<2Bo*^=)q{? zwxxM>9AbeZ)qpI`;O9`1*N3F7VQoaXHsZI=crP1g@5Q9zFNOW7JL-(>`OV~s^2)?o z5iyZLABQDW+&|WO`?|MgC1P)cY$-8Y<=*?mQjr<3sFTxApaWZrcgOF7&!*-y z9sV9?)t}J*-S@;ar54pg(yu!xoR$8*L&QYo(e2{fr;&IKH{KmPT*nJ}X(LL9WTiv; z-(IfKJ5(A_ut@sEGE?*9J(CZCL4L?Ab)IWh2*X`lW|Kz{_1@b}A_C=3`G?EXySbu0 z(>+~bXL*F%j>QPN;wv89mEvun;H^MAmLRT%# zo8Z=2bU`}zio{t^wvC7M>#;)kZibhlqJ^if|dcuNqf zNQmAs#ie>_!@^!`(lovjsmKF)-*9ZReHDP7|6a^p^-;4ke8Exg6RxGv1IO6cMn0?; zic~)lU-S{?e$=$Bpvl64co z^KBHesd;*RR+Us)L51a{)}&cp!hcA&_+(k6#_eWM&*#F&EMhcq^fW=v-ZCz^K_LQ> z<;TCuS8{ujmhO$Izlt|?i76C_W2HHwK<96aerQonluG_|fSCEUZxmJ(g)}^tq%roLn zTEo0slv6Ft#BA%KIpS9MyHv*4S$eUK_n&U4emA6{prJWEfUq6vbTojl?FwzBruc9# zGQe*zVCYrAv7}dog7gG}6+MC2o0I+$q^^**&Ro=o>!~|Dmw8H>m?y`s+BEyRcpw$k z8vowo&2rqu(G2fHHuZ`+ic|t=`=1H?BWCxTzh74u4#*t)n6a7A`^y6;gdAuKG&e`D9MZj6~fmazC5+UFvf!b>qnWM@u+yc~_3^CT;Mp$ZQxi zPpI|J97}#jsB{~m#|uNUU+uj}3l}rEC|$KhzQWF5>@_|R|9ZqYrwlo3BG)+0H~O&x zv7?9Hxl!t1cS8`^hjOSXDz8wZbsl;bq(1vd5MdHI{Gj0BgMwe3yo4RRofX$J@9f~% zBz_ydcUP0hpf^ad*MG7V9y$%+&sA79GYYo2j?o2;foBHUXgA_e6b+sEJGwhb@*hM1 z7WbPR_L+iF$9KfF@K4&ex=J5r%cZ`L8f#})x%G*tJt>ZN$FA(Rdh)0eg*-5x3W)b&;i(=Dk9knP zQa`K87s|mojZO|9H2CTgc_o}IA#h@0OMIcKr+U}u_@jVQ<_5l#Y*0(84Jaf7;4dQ?(eR8lLJQ)c4SPF)#x@Y+}K|E5LUpxQt_VH(?;lkWL- zrJ8Mp=O7uUwmxtkpGZNoRYu;?;j+2e35J#>Lo&AzUA)Sg`#HOIFRZqsohC6k?pe}| zg7U-L`QJGJ0(I5iVcYyr+x#6)4$kWw3j)Gk+7ED~)DM;!^FtZ)XTl$sJ2%iHdmOBl z07iqzY=KF9?XIvdAt9l+@~?!ldh-$u@Oy|~4fjq$+|YT&MMt9%GB+^O`+BYP0!38P zscC+wDLS9=k&0%7O7HMLv;3V-cFqp=g;Ue-%g9Ko+yt+SNIy}Xhj;U5F2;!7JG$@d zzTl^807x46-p#jAD(>IU?oNt}PD-G-P}BKFv${dU)}-in)#F*NrqXYZu62rJ)pkC9 z7MFTW%_I5c`zPh4eGjgdM|z`p^={SS;w5zR0hws9^JdGsV>wTM5GGf$ro;XD*C*Vu9&S~I&1TyX#0+KLZhNxZ#jVP zMXJGML@={1QZ!IPxl^%0ViPKGDE4jc?9*3uB9XWbS@2v0n zz>#uRb*uB??TdRpf-jaoqeRG-Hc8D?R&vG+07(BZ4|~}ny+liS*h+bV^i_!URj&P{ z|IB>C74E(N7Ni#r_hMPBVi;Ku$r|-}xc?IU>;s!zJBd`j4DD;|uY*4kGA0d{q|Sd{ zeR*vf4WG)06(MU`9Wf_aF?g%4Nmi$u3-_(2WIcbnGg3v2DCFda;^Z>M>lpxe`Y~J- zoYyJraujdfvX45Vt|aQVt)IHdM zv<)|p!s_O?UP{dmJiVzf^_`A1+;`GM$5u3u_}is9ajNb zXaC#1BIzE}xA?>!el0}AnI@ff@pHFr;X*sXcE>J#IA%3cI@b;XW zi86KVZoSV$O6;j3#YXtHtuPE7TAusEw9~58N{#f`(vwu~>6_ltEpvM%>-5U2yhy!P z>&&GC?wZax+OiIN^mcg>8(S{V38e}#CH9Kr7ytma?kk+vuh_K}?{D@xsecgZ&01lX z3+6;nkY^s~E8W&#C*-Tg+pJNe@gUSt@lkbt+(xNM1F7+}mL{K7yP-p+HfYE%j*{h3 zpcP(K3(ORjWz6{f;EiA{B~$X=8zT{p=vAo63n9v-k7h+;*UmIViYX~5*!b8D2W}d4pGm9lH?26)uw~X>Wb?TZe>nvI zH;dBhdm=`CB}V;&`3RD45zeA9#jG*KhfeeZue%#R#~Y$9nG`&{o+o{|r;SW&(! z@s%cn+9;{IaA}ZwmjZwFm+_V9NE~Hp1MK{h>S9Cb+F(*Hnt!e(jlI9!3&RS5c`rQMfFoaGI7-;gZ zj5%uRD{1O)Mk7LNG$nPpuH<7`2DxBf1m*ECCtmwAXO`4W>=)g?-RkcyUZTmTGg3pX z*5l8A?@sSMqj=q43uRWV z)uyBVs8O3cTM7(DX6d39hJ6xz+j>rO$SV}Lp;Mf^5S%=ETp=WpQ0egbWEJ<>kU*5^ z!41pnHD{CDIvIB;TMy@Rs;5@%e-}M>>5(V2*4%xF7Ago4D)1*rHBd@5SjRUsPPZio zq%80wZXg>9G_qeMrYk31lXx8wOSD2x-aEGw>y(js_YNnqSE4h4LTB&15p&RV@Q=yASwIiTFP2@6GK&vJTCABOKA4oo zRMuWVhkOdTSHef=r}^YTZ*!^*3U7vW-qT-)|PQ(0hpLBD?v&8KZ@eIX#c=r6hxBRRoPP;<=z^39DwwPo0CwN*PqOaB|4Q0{`1@4V2zuOEMK$ z>|0toSa)mPuVGepHzXx8DJ82Y|3oIsz^#IVzCzJ9&}Rgy8yoTT8nqtVUY=;UGdfh} zKL8J!(&sj2S1w1{v`XsHPPiC8X~yNr?3VA~d#+$#!fIcldWGC1t_I4J(jt`7TIe%F zelRPQcPL?T(n76FM=xFG*IY9TFi~PO-1XF+ei~i;V#WMk|I&xoc==*u^0kG-;VcPT zX#3~4nxd^DoJz??cRW(BggKhixg#&hdHEZX=!P)JO{8WVgi(h0ICTXR>W^Jp&*=AD zUv)}OlKw{2Yv;AbM@BISo!%`Bxm!q&>wqM3C>=hZtiDgX*(SIuWBSuF;A^iZ6)vt0 zjoh0gMO8x;JuOYvqLRHi(deXO(*X2DO1*K?JgzpOzRuXo^ynwILx~M6BFYoYGx&6V zeTUeg#^=8DE6El(cE9}jB@vB-m2L^2CoNqKl3KVO;?EW8OIYhm zRBe-?xM?V_L5HwGM{J5nDxKlp*EoEy4Ck${Y!^FZ#=dLxXb`Oo-DNo({m_#U=$;a2 z$!EVYo^od0yrFiYm(oWApf8L57(1z0PxTqPoyn=`-X25fw&pPfQAL|I%}j~7q{s}& zcGv$xn9rJKHH>fe=Ax^bjmnUDt$uq0ICLHC73tK#EzzAJ(`&#@(VZd{P(3Ov&uN%q zB-c8J)@>>#>U6#7xAv#0dv@XuO&$_@?I&NnsVgY(isq)|+Urov=pX?WP6O*4B{APv zE(HpC(%+BtULBWTHIyU1OC}3^$PQx)kgo7Fu%E2&ckK{WY0~di zd!}#7_u?Qlm~_zo*@E_KE2YRD!+Z&@_k-QFvvE#$!UJW-_gA`qqZ!}kaap5V#h*Wo z`YvvfmREn$ajPPMKx*X1p%zt#Iw-}y^^nm?GM$5zO^tp}YGn7DVYO*N%_SMNYo%Hi zqAi9gXxW-O$MscbW*G`+hn+PJ=o8?Kw`yZj!ro|iE>Iq=_v1_ON!_9G?&e6~aivw# zq05v`UMF)zKR{>Dmfc_U^gW!VFeR0_Jyy&3 zKwKxr{b8ASL!kCxHgfvRiCVwEbk6b#$Cu#WdnZj?_tdPb2|To`cjCr_i+K$`Ezo&0 zEmPbgJ$W}q5mNRN9sH5PtBkGl75#=gZ&gDy6;V}~9Qxs|857rP_*Sm4*M?Ww@RDnA zNsdU<7j&bg_CS4`%RY|v-m?5Z#6l-s)U7M0@kI0{ej;&RPE)?-l?kh>r=|nnNowWW z;3WB^(eBy4l5%(Hr7>y|qOz`3Hd|n7w%{WCX0%pGP(VX2r&IE=ZgJVI^=|Uz53j>c z8C{5^#_tkz(*rSL`I5$Z0}mBlmM_(YQ8-_hnd9|a8u+zbrHM_dg`t_wJh zDY{5uOo}k|F(=;XFxMmi6fE1Res1n%!1CB&YwKpIfQQ@|rAL zJ-eAjyJSPZ>y6jF(}W*yf7CcFfm0q8H&{u57!36wJ z5q64DLxT|&VOJVgr4HHN(;phEk+VIgZ`3bgRtQacSk#%L8mAD=bLy?d)>_$j+_FE~ z(6{@P^)9{7Tc=UquVa!0mxiyUYC0+?@C%<%CmLDK7#)aw3S~G9cVMZRYBV{kwa9OV zt%j}WKEIbn3Joih4mPDoK4Z0}KcFt8ysn)`DTW^Ah)7kzBTXXi%+&S0fKiU#rRrf^|L7mcA{j}XW8CAEDLxNRHP~s6Ws>o-E}CuYQ`y- zSE|t;9M(!sfq2V;Cp}jo-|wkq)ry0nTT9v;AJSA$)3iAv(p0b5uU@^U948SS$Xl)! z#}qv?sUOP%kOeDf?l?B$Y<`Sr2QfbHnqOSJYty>u{e+c;-Y2lwP41URHXKGqC5Pdd zarZUFSG{UK^_*#}%$_MX)@qqGuaVzh*MiWgy%62HsL4T;qNQ5iYSGoi zf@ziU`Xy$g(&2?CI$9e^G_!cchl8}@+I0P%+P+G{_|A=~>j%~n(T|&*;U%LDy-W3F zeg>l*ile1-bGofDS4BT$t6m9cz!`@&Jx7*aKF`8}zLIp>Ib6#+}%(tto&GPODNh*a&POo#77eyVZX&s6Vb zQ+}7_s1w^Lankr*U_=zCD))6ejm%A;KH0cOvGcVkNU-uz-EAP{Fyl1`Ct>@`VSESi zXjTsT_0EXavBZb)>@=t@y2`3QjjmO@j*A#}yv}*Jx1?|~*VMvVacfQ4J)yLz7e625 zKue2;csqXA3s;RJizZV$H;3MiK59Ru6m0CvRuK0H6gh|^;@x5O^s<;E+p(km%qkAyM@QG{wZ{<%5f zE16wrRk)BQ$^15&^#;Alw32%GLA22|^(0ljBG(`Srx8cN^dtki-)5xWH8>grgNBli zIxesm=!Ykv(^Z?CG(3zAQu4*kNHM)au0NKcSlyag^F^cfJ!D@~YviCIgL=H`=@lme z#E2t-3cXUKs5Ilb*(y-yz)G8_CwSHQlsvX)bO0`D)KRd%mO2u1oRv*^Wd(D8W5lKA z?Q;DIUi@f8h4)$V#pf+wr<>k>VMowi32{bW*q=GrH>la7W*k=G#x0E8>A_Zwoxwq zBuifr|ryU-Utn52anF<<6JQ#wHaQoG{3%cH+q+(qS?IYH=bk` zVp3N6uLKKZh*Y;08Q=NvqKoPAr_5L;hftVQcdg4Ne|Kv>027{Rd2NweXr_AD=SIhJ$z#Ry=746JBjv+PYZj zd&%7$xj*G*yns>gK${R|2~c+2nkT#WN}Dk2)$L~}gWE(O$Jp;=_1O20B&H~NS3h}g z(q29GJJG*r!hQIA8G}jIqhgmIv5j36KG7%o6Lh&1n*#6{s_(&NvkB9ERZfbl(MNBV zjoWw99$gyndi(@mAQJxGSpH?rDYAQrgVUC`WyM2-*KT5+ax)xFX2yEnsGk&c8&!A) z#jj`BQ1kuq?2q5TZt*Mi4&q>X8GGAoghhf|{1hMc{3*Wrlg%Q7JL&~p90o7DD72!< z!wdc~L9byto08XoRbL5!($|6cUkOHvH!Ve1$=3Ik&LXF6SG+iN_wEO+fh$Io%7@>La#byWw@Jjn_T*_>*4w*{q28_Hf76vDWrUSg6>eKsTyqnl z_DTC^)iwiiKo$)^Bqq$GicS7c)Qm?}p2OA8%;q1_>>U0wcj$fhI3ZlaSSgQn7j)WW z*L`;xhWs3j?3?@5ydbCgl#-d&!8&H3qK2~&HVTW6`*M8S%`bNc-w9@wC1kiA(lB|{ zqVV{JK+NQ66pAe47l>Z-Tm}J#Xir@$XuSdFHlDn|El=RN4e2e^0%t3{9W`{Ea-Jk8 z@?C-GPoBll)dTcSmjhj3N0;hxbO%E;jpxvHaq#&=g}`;1nk&oC4cGw#j~tAOLg}1& z*!Wdnh18|q4<4%K9eLl%0s8KhJ>FV*gZjk&E{1?BPvEgNY4|Xj*xHF2l1>V<;sbyGCfFa5_BK8o^6V$R&BV z#s_Hw(UC7>1l1C#&PZGkHs6`e!fr~9#kP6qPkZnrB26@;SOiu2#*}iU{p+oVcCD`x z$~nn+b(f02Q8j^1o6AL>NW>m_4zG2a0jQMmY`{~!91-Oc$=zu)WF1($ZZ#Fd!H8I6 z;3M(%wEag!@lQtIQ#FK=o>bnsH?)>XQHx)A+Dp_zOFW{1@-^|wbUJPX6%DC8R9qMb zF9I7Cz55yNqK?yK_&Z`Yo2aW7dVpR!rXGpz7*rztV7YEBChv4#GU=3&Tj z@);5R_~^*3fl8B9#j28O?xuR|So8>T+e8dj>GKNBpz1K_jbMp_v>+c%thz`CvzXot z4FA%0vH~JBU?V%ZwCkV9gr;j0YF<+u=`a)Fd_mMaF0x#7`1URU>bKA!FLm5OQJ(gx z!<~bg>=>D2GNv6eiQl6xe+#(z5D_ipG@G*RkOTgxskBL~Ze)~E+r)BOP(`T}w1(}o>-O~Pv&+>v+ zY^R&1SHlO`Dipq_N)^R}0mLYRtaAoqr5TIJ(X;q~qQ@(#gAkDdAs-<7a@{uj^4S_g zUhq)d^p+^_Dk}?V!pReFEd72aHwxUb zRxZ@CvBl)-MJIn26$Ct2BZK5A$%RjbbOe1Iul*_HGAPM}cg9v^yv}wdOe4dx?tQp0 z@HS&mX%Gxi;B-MeePJK}n~pf}(rCVmOQ@%v%O1C%?~fbteLO}$E1MUc`fJ6$>38DL zvAkqQT#&r>7Gud;c8k#q#?cMI0{N=#%L;17y@)Aa?q-Mh4}c_G{vQ1dofg6H=zoL! ze4TbWrL<2ip45+Hk_5D~46Ci#7}5LqCz6V6pVnx_1_AK@BoAyj#t(#m9ogr@7ptKCvBcG+v3I1YrW)3 z;L;VccM5Ty+0?#xt9XGA6neiZxa~#zKVc;wxJA$2W$-HWej;!@47Q$ueNGk#o8vH} z!0+~ILl>W7z>X-Dr>qxti{8FP&fJS>9KN8glk3}JC)~)Z;w}5Rn#9|^6D?}wHA=M? z7oZo5t~k!9)5!mNT$AH8PW_JH_hNC4T6J9&ptFnpUZ_%~yU208a=E@|YOpSjJ_?o& zT*vI-OXkYok15Ji!i6(h9BW09-Oi;%NG|GOp`n)L-^$SDV~67p&$!1qc;p1~9GlsB z(o35)^Ok0>8ZNpI5}pi&8YUsx5k-et^hpnXyCgzVvwNY|?=jpSq|&Eq+ZU!e7Owv0 zWrMAaXT9;-&=ekeb@IJg!&FHI2bkJ`g0Tr-y~k{%v!X7frAaBiFf_jXL!XGj(ilsH z(&BOLjWf5$-n8r*A*N>YQAy$nL3CJMrFndr%$>X%o&WtqM|yMmr$_JAbU}|-CwJ%p zOAX=gB^Tr~88ZEgsAWWQE8iru?4K1niazaanQWOpomwU>jX%dxDm#^j*by{<%0s2o z*k*d;)(}eH1HxrB?Y-b4cqf8qoZGiga`}5e z>^TVfH*TP+LPI^-8E>hh_LdAtwg(aE*)7ueND$FKJ|D;1Z5>bBZOG)>@UId?^If=t zf51Cr>eKRCQa+3Ki@?Nhg0xY1hroGRlW;G%prV_MHttv|tryDNs-QJ3$*h&BQl`=?&r>j?3806uCDFl3JSK`C z9aV}`IiqBk>wbnj)%`>lRD+xhH46K5f1n>P$Ef{VuuUIb+E5LBNlJyvCw_WNv z{DcTe0??xveLkwwL7})0!vPdQ>6~xHy#;t z+l74)ZmCwD@?YzRkiI`x{GBoHR&Ke!e)zT&xYlcHl{~sKdkMU!3PgpSYYl$h7rrPJ z_*BF;r>Jd%RMk?!>QlmST5GOY8tPtOs>_2qQI+Cd-FmpC#<_GsLpBL z;t#%hBOf!qaQ;r+H7(nEQb}(mB)Wda?$w*f=Y4jmD;7w3hh#oDB1l?q&cTj;6& z<`XYu=4H702;#Qwsu`P+)(1MRqnF_eqSHO)~KL3?$rOgB5`IaJjJ0QL*U0b zx!x8u?R5Wga$PK_xji=`OA8u(cF3O1-pFx{&YI)+H`iB%hyD{6T~Kfr8tcCJFRZL4 z?wQJgXffbRk4a}OifK`paXvouzvn-4%1h&roDS=Bk#vVvK^a!kn&Llf$XJI z1?(6;1pSdy1MM-T;zD?a1&GF{j6jpM9>A9>yAzldx2?L3RYUilt)FnB!+EgF3~Eha zUu?Q9Yz?pI{NxwC6P{MhANpFbO8E&(xBSOe{I0W5=c}}Dg~>uu>MdfhYGU^CAEvsy z&`37IV=!#NC+OW9*WoYC8`OmEAELEYS4MB@ueYyxTVvRq*wuyR%wfn)+-5mHPyuas zLMxOl?ltA8UHI~(_t$RFSmEzr@<;9aVXtX&{{nfD%>?5@k3;wRhC`te?8hzLT@!P+ zZMhcL9xOp5rCJY4uOHp|7Mxc0KV`fwP|p%dVEBk~$*)!^lfR~=BrZiE&d$D*(XPMg zusb|j)~Ih;8H5T1QV4D8`!}4;mw~!p^~`bkRnES4fo8^uIDMn(zs*!2z1BY)DA~`E zjhp%`Nm`G+-@ch=C&9YSroWW)J(}5RFxSa;ujfcD4K;G#a<`~|B&ceGb%|3Rr${Z7 znj*iFPkPRXroEB13RHzZ7R0O@YSKi-hh%r9v8w1@DYOx*o7>ks@x(3z!! zcL!LVPU}>w2xb2e3Vn?3IzD+*Gj$Sb72>|Mu6gHHkmc*{)w44T@qF}6RWO)H!i3J! zI)ysrv0HYvH`}Cdq#CO^yNy^5acTE@XhX)bl08-Ft_ry#l)1K81p*@t|5V~icOJW8 za<2-av~NMCS1`24O(b&+9sob$3#%Pw0)+Ng2CCwJh=(QHF|NQ!@SjKAz=0Ff*I91s zY`2l~&L5o=21dw~>?X8n;=)peJ}Y+~i@vUz5)F{6*pIC!lsT;9>O9VPJt8g)cfZ_rdeXvlUu;hd70|W?4)(iQsCq~DyQwZ_SGuE5)A+YC z3a=TCzpA{pL=GJS1EE9LFDG7!-Gq4*y@n!)0i5cVD!w)B5<)nQsfXQqiROB(J4>-8 zH*3S`va=LedXPW%pAF41*XK|q6;cBMURPC=F{4D7?D?O*izT%kCmncky9sOUhdfO+ zSuVVc>I}_)BUF-K=;M2>W~+8fapBaa24PUWu$1tY9^$SZAd&YsR4*+B-6p){9|t~m zCN4#R3~zV07X>leiVtF}LU*l1gZX))PEA#Z*RU zC4cnSQg-qik}|;Zb@n0-hjG%~F9`xA+SW~WTSD|O_DbAqj`nMgy<=nSgVuWvfp4;b zvzV-SUQsg{0MyrBd!gkkdrVa}sl~*lk+N|>{Z$FM!A_DDt= zc(t)y@xj$i-c3-IXDWPrQrLkSHd7H%QBs0jO`L9yyZN2?gZU>)2eNGgTG~X-su~R^ zF7jn}!0vlHjn>ug=1Gh9#Xy(&17}JcN>;Tcre;4_vmg4q8Z)ap?1S5K(MM2;@uC@P zyBX^oNy)_xS(UxuM_4}?)}Nv`OL00k&#NgEDA>hfVy@9WdI`JUenoGtAu-xyCTDgz zDQ)fUj0&E|DqKXS-K258S!cOErJo3&Q9X46%&~VlO%+JlUA&+NG0NI&@H@rXW-SoX1v6=bMafj?>k z9MukGN*>vF(z+EtX-9Cs5b(mzI#rw-*=D7e$%XO)sK{wBL21U1J8OF;=K8#p0{vit z{@eZH#xF#hwVBoP_N{5cpa`QX_vc3(8c7y6h56$ zNb|VCC%@BpWja53zUGLHW`L%-W)xQx7 z7hmNtk1Bd!(qXkX6k{K;u_ZM)*>=6HY{2}}ux>!2LJ;G$Ww!ZQ}k{5mSORD7xM`H3;@6B(TZcodu-_qp4}+ce`lo1&YNwTkG~!s?lm7^Q=4+h6)SqPI891L%t-#%l*dTkMaH zhg78S*$q#bR=)4xs!>+hn(C^h2A{?We*Evt1N{cR0;LPNDl?e~CI zy+xl^73nPndNbTzK)*Fz&Hb^uFuz$tm2In6n?|zVnry+^I-$rwh-ka(Lvj58pH4#@^@61|uWxq;mE$8mbLE zgLB9t67?5N1B-;W^Qg4O-~7FukwTa@dVegTG02NJ#{rVpdQo!&>+rcOQC(&Ph#hqf z7!o4-Yf14tzG3BLlah+RH;$Yn@W)rv+Bk2`zWg_`cdI@cxn7 zM}fPez<`UdKk%5(u3O`tNH-Pw4?!x1g+M!tz%7+UHnbc*9iwVHm>uHaFD6MQgETo-6(3FDK-u!w!n{e zTq&jlKThj6tGWuhvHybqB2rt7$?oP$)h?^;B&c}l;cgP(E5vJ>$S<~G9A~y#XCI<( z-1Ti#$c%IN|L$VJ9P>{%w3USaipFyuFy$5fmeS%J7Ao-%%_T5xY-i!bme?kPCb|Uy z93JBKaN|wX=UE%1AK}4rR$1g({-Lrt&tp|=>J~+NK6~b3V0ZaU6jR#ah4dYh_Ct!c zW4A|F+Vv{v@8$|V#wIv)AXR;z=Vnw{c9^|UYhvr$JSVmT!sslb_Yn5Fgf;SGH);D! z%8$x$9sT~+1e?Xiim(fty$+~T+{YRqog>k{N358Bmy>B<+7u9ojS5}1 ze60o{sl35d3B~zM5D~&K!+jCLabI4 zrq4I`jhZS}y<*~m$1Rs>K+Eq+5EB0zO#V>Z-<8hj(p;D9s<<%UmdK3~gWgR5EAtarGruavqovmykT)0WFymdhzy4Dp@crOJ|+ z4+p4muZBOenB>~5Y5&ndj8dzwYnVr4v6*PZkRcR!z_u*V7u~Orx($$RS)NI7ap$7h z&z6?*kMZJes^e z5F#SFtg~HB)Rs@hKybck;k5bOL&M{PoEzP_rqugNskbde+WV?Klm~!sCG^hYV-dok zgk<5C2;?|2T?`F942`eoaXk64dwsmWsv0%p%7yk{tHg27vMhvJT+NcR4fJ6TP=(*d zH@u;u|J4aE?t7KhV<4^y82d;W*?q#((8JR>ms-a*FW(-`%hO!MQ))6}H?dL4-U(3c zxW(}^B)La$D_V(dsS({G@$WWCC<&pN`k4f8sF5tY#Q`&OcYy;ksZ8tjkNVBDyeY(I$hm zvOsYVukaBD$gM2kR+dbppEM}&FBEz;SM8pTUJZXy%k8QZ2BM`TF3%p8%{V92c`N(br&b4jP<4wcnf-UOSv<2x zk;;W;_&nIoEMR9AC~XBHoSzUiJK}Y2L;#Ny3)qPzBKEDOUcCAb%i-=ZWK7IA40_jI zUU>J75fZw)RMCiee8?LFc4!2JHOuQRW$+*&z=}^MZwZ^e51H#2DCUHb2Vtw`dK_4| z+`1*@R)+uk`AFQ4+!fqugm{PL&|$#wnF!X_6zj8f4{}v1qsLC>e5(r-K0gV-!?G~ zRfuQ{(}T{Sj%;Pm>U{f#B)0{(U&P(!PG^#y!*j<1 zxnsH8cSy(PNdI#uFC>?*63daQRyu?STQEDhEw4%w{7A_Pp=7;6%J#^#Yig>#F%izS znYD!2+zPN+C!lkV0TqV10?DGLA3@*aR1>i>9&BQ*Qw2~aD?6~lsS!P2Y(v9n@5U>R2I9z_|1V!!v}wn|OX zQUc8M81s_+<V*K`YToH`M8%cz4n-iERaMNiKt$@nBLPLMim`K75g#0mEmUu9kgBZ&@J>g zT|R&@1*#5Rg1fJB)k)yZ(U&J5zR%-6cDi*WS6)%P$VyH2B7LO4r6qRP+rv-KDWWaSYcWRHBn z3^`y9@$4+kLKkb|IF0xO4r0!dP{qdF?ppxR{VBwSz@^Mr?M@!MFXJtMVIt0EwrH)| zy+A4*z8)GF+c=70*r+D{nC6%fCk4yw;#)&dik52}Po840zt-Vv8`?nk4YYhUgM77! zUOE=Y1oIk_oCkP<*+cnAN2cl^#yF)v`S37OtE|)i&pv)z(6|36^S`XFT^0SA0R1`& zy7`z)w!BWaG#za|C@BaxzVThVHX()$N!2LKEQeS*8S;kq`$K|3PJ8@^;et5H@8PZA z5x(-Wo0c1k8Vm1kZg1&9g0l7Krsb_a{*eAMRW}e;k68Aj+w|LTPSWpg^-t8>N!3y} zXj7AO7U$BUE?3|h-^v@|AIWo~0I0tyfO=K#sOqwz<8fJw7bx&CIz{+|ijnhjuJVLr zXtpOKwHjEV;WyES^+dS^kGJHFZQ2tf?6{Bx2Sf$2h`rKw|u6IrZ2)PrVs-_#Vd-}NX0o`S#HK;^U8 zH1uEk2#Fq>7Ihqi%k#lX#XTJzDjMfqX?M6%0Vf6%E1f`oA2s}a=Gni#DHS2ovxeVN z;+EruGpaCC(I%Z8mr=ZR0$6Ev!WfxcZc#Opq`O9LXE$36TuO1fSQo^pN&_|wfEadm zZ4+&V5zLLUAHDt$2PV41F2=qU!z?RwH5wfub7$=l{DHspCnN%|rk@LezF>yEV2*lw z8eF;!}<9Kk2!8paFfX}jmh3}2-#qfPu+kKh+ zCcrbku3i7<`g={i1GYH$zNp8)H!k)Z=8h}^dQhSCu%D@vkl@bcZkMh2TS_ag+vZr1 znuD8W-MDqiCK=yfS5o{O03XB78KSOeODTKbdh-EsQwT{ZdnWvpTSeflqT3W3r-h>d z(d*g|YHC0Q<6I>rYNQ+ZbukKh;zgyJkULRGZ?NN!7&(lwmfZ||PG;CEW>Y*@rNqhF zS}D6NXDcvvaZeSZn%XTnGax_9SKz0OjAIRYQ{YrT;nq5m(sjsjB8Vu4ega`uS>n zVvT;E!L@r9H$^%(7YM9E@JZ3T98#hFgUmC=xvJP7hLfDaX)cKj&$1B&jMR%0zs0X5LGW)_p*0NYM4!w4^JwCt2VOK;XbO3^Ifv& zWp97GaerK5AtY0%1RodkFVE0)0j(XC?|*fv>o05Iwp2Eo>90O#nQYx_25s5g+d+Qf4l&0*By zbBFw;n>FVmCWE8IJ1Wf2MYxnzZIo5-Uw&rB{LIXuP_ahY#!{8UD~XN!8YtLO2}t(h z*!j5lr{U>x$04m(A5H7P_m2=VVvA!;vzjLKbVQt~!_rx@d`2>MwzLT@H}!?1(Sax& z=WKdq=4(oQPYz^@N|7qJyuH#xEH@kCd9TORLF>#*pI{%Xux~5Qwt*8fP4!D=MnD|{ z5=WLM13kPVV>F zr9>P%LOqBEw1>kvxdQ6q?+A11!DiA0RbVwz?cs<`>Qht*FK1liAWcvKVNwKvg<=?m z^s!l8-hy3>D0%HzDc1SaOPA(vw<*o-zlZhh7)Cm)NWo;C{uNqxYX3o~ZQAtyqYo)Y za-z6|2V+U#j8FmenPL!Xv(`_Jk}pO|-w0BnWV8#THCB^BUo+&BHr&vB1jIKZr|%)< zJ&s5xu-YQkD3l~W<9-f@;6%ruZ+`TcT0y)U6+>}(8wK(zF6~djqy7cEQJID4=DV+c z64kIFTzb!QC-H0heSLQSAQgJ7$QZMMFJA$ERgSp5@ut9x3R#f)Md7cm8UbS^LJ8g6 zeM4?77lQeP&%}{SpJb9qsm$R;SQ~Ij6u$|om#TLX9PnbS#@=0=n28L*SZ=)t{#fji zx9}`1rU>^u?i3gEzW=Ng<%oSLXyX;DdXHn|D01MAKlU-pQ|o={&-=^AR-4ah8uik7 zMQqDF|KxLB*nl6w>&iw&%x<@0rXb>O8Qun$h$ zjM7VbxIi8oSh228Ov)9UyW;cM93T^RQDs6113?~@L&8LHgeBcSQp7O%f;>JCE`&db zOod!-omCZ^uwHm}vX*@b_SphwiWhOsBI2O%oY$c}3!X!Q=3;f<=w%0I1EjWXkdfJA(}tq<|C}^wR=tOA5s%o6sxdosaq8eRo#3)ffbx>Y5b$+G*}6(f@;(E|1KSkyE^zSA zQ3f`-Kj`>vI?ghQK?~13{&98CLem6NkEtoBy-^?aO7o2m*9r|jX#cGTdXy}Gbn{i+ z;&XMAp_mfe7U?ezIF)^j>fNZ9_d0;BdHqf)34HIZK55qq^q%SW6pW8|OnNs5c zuaI1#>s~3vAB?~}no*V#O*ein&J^TwNu-YIq4B2d3*?x4C-VjFJGGk6xMNQ8v=!S_ zG3+&XwajaaQ6p}kredkHymtrRL&uAP^*os;PQF=`zoxOQ!2I5?Q9~Ve0|jXC*d*Ym z;n1&5xVfG*(Wg!%0=R)5S(W7m)k9Lq*Y`rpNihG6)op}mOINx?B$$iz;CEmogE0s@ z;i%>FO}G67Ym!13#f6&_;^D#s z_c)A;6Ef6KNd`HGaP?!Qsa-@ce^{M>L9ji?=svXZhaD$^M{a0&Ik$b9WxpZ zlYAN5K(^!EKMKBNn3UvK?eHHd{6Rv)r1z)!iL*!ucM!^~h||&;L91+4=tjA#h4o{8 zc}?Pzf961Z0tZy-i`>;W6Aoj^4^&O*d-dvIdAC@*z_&7|eaH}DL3i;R&gZd=2Y5itPn+|W$v*gYEvts4rR^au#uif6% zl?L;T1R0B9g`1|qJKPE7xPY^qw5wv7WC1tTM7dCUp5^$yukY7zJM}Pl6%EBv3|mIf zFnOQd24r{r#C#svi`XwHhHbz2qkHr$copSgMqp)9g|8IXbCo@1gd2tG@(Z0W z@H-SFaClUG?M%!D&|6`JL{sDSG2gU}YXp0{&pUc#6;&G2csLB@wb_!IOtf`NdK;qT^B^F@H$NW&!^NcDaBes-gX!8IdR z^GGrE!TdiXvp<1cB32o4KY`pL7}wo^#vWB7r~3M>_EHfWMYnowW=tUovU9;qbIkz$ z$s55$`M0z4;{?{DzOxp;oqxN6atI7D8%Xt?`F5v`mz>ReDVL3HKq!^{3X$5*q_c4B zh+EHN@2_)mH0OHWm}tCRIJPj+mImnl6=FxR19TrQa6DC{U?1CMY8LA~WHAm*Q8q|{ z?=>d!ogB^~CN66bCU-;3qb^gLK&!Bwh3kD=hX|0;zmvqHKj9Dpl5{eQ=`N6KI0cex zIOO=fQp;Yc{$1|yiQP^;nAos`>kwxOVm0Om&k+4*2_v(9XNbc^9<#ySMQefRdj3~X zX4n3YEKp27E#`yY1fU0DBh@oiGAK?IG$ z+3X*&oO_KEH*Sca>UtLx+;v4! z@x?hUPt63CN&9bVPZ%T$2EO4i`{5^% zIp^gWULkj}kM7J`4lzVqlxu`E?k6B|_O4(Z(g@;O_(*6iu@HbAX4lAjPF@So32pZa z(}E3lu82iyg>`kldmq$wKyh|s<(*1&)9q5LQ9k?xId1CJTX8lAgNa;PzL~P*m%KJa zjED(*76Te8y=T?gN%n$>_d$W7-7U6|z?D^q6vPnL-YKpKvtbA~|2r+fU=n6t&CB1T z*KMguDs-;beXId%H8uOwEEm%m+Ru?*rnckq$OAz)D)9*X;^$fr z`3Pw9=TSgiOR#qgp_0%NtkkQE+oWL?9W<9Jl&&8UQU=1)ZP#zw->0zcFl1yQM+)c! zVT5*n(uV}Lt@gf@5DaWHuwi8B`l>F-*Av|RNnOxw8CRqgJiV?1vNY^S9v5dHi%h%& zAam%OZS;7PgIU(n)WMEtCmi;I2!+p976tl*Tkrr0OYKwgn>9&&-^1eU3Ek37a{Y-!~^4 z6qN__auXv($o>}q|Al;%yxgtJ(;XzF$$GYyZCD&BkdNH@*g`1~Fn1jhXz%S-E;#JW z{Q?=^FeU=@J;_BEJ6+G(@>N~=(t8I~ExdF+drOYM=f?jF>_6FC`u3Jfi7e?kh+Q%c z7m*I>%7RU~S5S!$;gICxAo4>vd-8lgkZ=*$Ady)wneOp*QPQ;Ci;NWV&%1!_0JKN-#f`I+Do4ewi}m34Klz&Vp0QxYNg zVQ;%2Cfq)74Sh@sX6o+P86!+%?$-YW>px;=kd;gTe49Q?U#7}lQN~L2Tm=0Z6X(OM zv2X)PQ?V|LZL7eo-21?0AXyiVG^xwJB~^31n!L=upcdb25Fa~Kd0)rsS^XhBp{T*_ z_CEDSNo2HWdA~AvXmBtZ%zt-Pv#{IzRGj=o{g5JOMQHVqJQW(p-I`om1*P3g*qeGB z@fHSj%i{MY0z&G~_b#T!-gr?t_NWlwY&BOsJY^!LIzL(oJO(fgD{b6epEBQ>k5SBB zi|lXZ?RYM7dndo6FuBRQ{fzrCWAS(uQkOcUdZVoKP4Mn&2}pIEx(PkBX};4wW{4rk z{g(I*t7yGmdiOWQbB!Q~j-u|ZQX&daa9ak5Ym=A^dK&rBVN zxC50UcQ*sO&W{bHiDfYF;7E;ICa`0}Z4_ZQiiIl=4BkB^!> z_n$$_QpGh^cUjlJa$$z|t;N?gtLBaxBbxw$->jhR4n z$J2$#qbc!v!qczWeVrk{BiWZpL%j#jd$Gh4h2B11=mu3?@I2j?wWgn~iV=JEsq(bm zM6H%ZEv1TLDI2{?cIrjmKs2(Qcw%pFqVzna)@ctq_{Jt8Ag zddP8ySz_zHM3v!k?02rZ&mtck=spB+Q}~fOYJ=xSEA$X3;*&I7)t?CzjpRsaR*u5G(UF}odW>Q<oRRd+~zdWE>96Xec zYLDSRTo3+vF|<_+4On!kc$9jhm?PoHF}-qiVL0d2EV;GQoxNE{WFk&~#@4@9(UWKL zf~}NimQu#Wxg@EpFw!z@uRiI+H?*)&4e( z_c33*i?sg1bBbaKjigTkdHyB_S~p;N9T7tNoR%DyE{v>X=eCzxQ+EMb@84`>bE>(s zhmvb9(huunLh#K1a>ADH1O=+mosQiT@;2;ai6Q+pOS3D(mbZD~QLD5qV^&Po_x7F`9r%l#4LZSVZe znB=dr_P?wJ_8Vm%>CR+e$)+MgdtK(?cjljrV!Y>7NL&A$?D5FS{nZb>7%}f1?)I!8 z3b%vV=_Qu53B)%vMcA@dt$oGWUDvm&k?gEg=Q(D3JTV8WqADQm;Zl>Uv6+5vs5Tc8 z*B)UQ_CizjBr*GdpUa3h{I93E;v}^=8YO7-sp4?55#3{pA(z7Z$it^EUv2NLbqp=-jR>bNrNW~slUC-aD-NQ?2;Q5hR3*R->vEYH zxGz6KY;4dwq=zzq%wum4-PdY+0V;tuT>+}07oOhm?LJ4RzmFzrbNh66Hb99mCu@JR zH{{#&-NId{Q6Gb%Zv8R^3A4|pK;G@oZ|qIzn>!P`ODIQAF_zsdUsc~NNd38ZPtw#C zJOo*FNVz6u5M~G-LSJ==zD}VP&I=x5T6Lg0sf)V)Cho}wPzw{`r-z0N$p#O}=I_B4 zaG;C0r}1B}{1tct6kGxnYIEFwg}bkNKUI6R@emds@H^AX6JHv^I-GtHXT{3{84Hc`%vDOJIN-A45f?VCGohLmu1^oj2<(& z2zublipS3cxI8BAQ~I1HLg?F+DnLc3w7l=3vt>sraPj8hlaf}+f}RUQJ?M-UN0YSP3tqzE{?}9*)tor>y3j$sJlKcI;-H1HLM(>2JN5 z1+g>&=;I+$oOxwKmmh;c&`8ctl-o1y1l7+sha2}?4*HPj+u&&$dS~AO$V{aXE3>x8Y&__h8f0)TOk! zVLBh;`1*TkJ-^9~(H2r^3$c8DhB^Xs(XvmdF(jgtE19UGXGq%CaFje5-z?OK{+xgU ztNIRZt%PX2*MhhDK~UYd@Ncxc0Z{{o@cu78=AY|nskGbIX%wxzsv<{zvAux(#ktA_ zl1c7;`<;FiysH<-xRN!Mc!IhjH@duNd}^t?7j{kYJhg3(T*0qh$ZX}-TKh=S)@|Of z;n$?Lw9^;qM0V^S^|t-@(vSK>i-h}=un6y4wKlrw))5TiHPSu)RU`uv{2(lteSO;X zJ<$`po6re*3Y*{&2%IMq5;^JUb3?ugT#AD>XT}5cRk8nw@XWsu_<$%r^hs-GsCbK2AMdr{AXDTg0*X%A-7LZuq3pZKD?Wj!KjH z!E9hJ2b>-LtTsp-p(g<@T0Ww6I!RxAIIgQR*Sv)aG&b??v;b`lgq(aYzLy;!`_>U@ z!u&|H6the%TaDYjj9+_UFS;vdFLkgBHkECD(zoob73(+}{}NL;+a=w;5~fl*KvLh} zlybS(Gk`cQcj0)AowxczAh9MjJQcC>J4@YLtk>R?w^^5X|w znKOreZdX#@{!cY1;WwQnF$qoP5bnrU@V0~aIUn^cDXv#Uu&{)9`m-655Fu(w44{u} z{a?$55~D;9qkIpeS`VXc52I-hqfHN^OAjN^BN-&=EZE&R*zXSN#c0`(4&RJWuX z6j}#`-aw&z5>Ph@Xq^P~Mgqzw33Zc%)=5HdB%yp#P&X-PofPy&3d$!9b(4nHNkf4f zX(*oz)J+CjCj-5af%3tiZZK#a40;2D^2tKoWTADk&>LAOpB&Up4q7J%y^(|R$wS@b zp>^`m8+j<-SE$=pXx&%n%~vR&0@O_bTBiWLQGoI(LfsUhb&Aj%MJS&V)J+Lmrv$xG zg7PUt-ISqq%Fr8SD4z<{4N!sBsX%X3pnR%OH&tkzD)dGb%BKc(Q-jv2L2uNceCkj) zb-DisGAPvB%fNR>o0q{~vVVe|G2pMh}=7vHxEb zfVsz5$38XT+cuq=dvA`8eM!Q%eL6Ld-VL3;c8ve;1^C~*hIC*1|CRv%v%M;NSGVw` z`2pGKwCM#>N~BkldtGF0{^xFTG;GTdbA^Pp_Z zrlHcQfFbN(SF=&!;a8h8$KZ@ZF8#`=CRsqslwSM)VDGJi;)>R7U)()NNRY;z;1;Zb zK;r~=cXxs=G-z;lch{gnLU6a>f#B{0f1SPWen;-9SMS{O@2#5EU(Z=af3uN>t_qkdNPGKTN+^GJB_Gs zG`9zwX7}8eU!JG>)-(c@J(g+n8VoSJLO*1StkXH2pV28bI(k4}XHNYs9{&=LC?_wD^ta_x0V)+x4RP@yVp>!-pg66mLsYp`8R|CA#l=0*4 z{lR6wBc}_mfjuB+wfOW1T+-4ZgZ+HG+J1Gv%3Z9R#QlBr`ZF(G4gt{}!SmsMhaAx+Q$yOe8(Fb|4 zUi2@6V2V4XwYVU$gFLKedgxMVwih+|3}q|XyabCzzmjnFJ_x8D%xSd>6MK_yi3xjm z>anv!-uYY>J5FE8N#}Z;XEHQFc9)d7wd@K%eaBy~(s8Ma47cCz4%@q(e-wvPwOCU%ia>Rwbn$9xghAjBYy6D;M3bKG z6oqno?sQd$dfsH?fzhp%Mo)O`FNx>CR}?+r9f)*dfv*5k+i860Opr*9k{WFao;g** za!(4o-LPIt``c5fuY>bN+q8I!Y`qjxxB0JI48*SF(me7Y?nKf`$tH|9KMyB^c*{2& z#8|2GhU0w3(;!#P#ils9UUN$`AkKu=^idcp9n#HaT0h-VEKF)^_SI`gBVZ{Kt+bNrKIGB4^y9 z@mdy%aE%pMQ0xnm;>BpF#nfKKp{wxe$%F|P6S#AF&8vzF#` zh@y}HR!Sg^o;AgpGZj*A$QZ;{m#EqCtgkC_zR~4(*2XzD^*Ka>2qto|y|e5Ev^QEc)5VN3lWHHzhBs74 zvZNiWib?`{z68$=rKr40{#j2(OLMfnlC*008r|%f0;SJ%4}}ZAArCoNoSk%OK%D4s zIr2VwwIPFvBn0_Y{&Lu!xtRfAAcJp`uNeAjm5Jp}I6n?Z2OuL0OHaEwD;AvIX76h3 z5Snmr88CzV=?hRk!hCic48@HywceOCubQEMy3X$(P3S=@v4T2>ORi%a=LjI7^|+tt zgM8>OktK+!C5S~bt3@Rf@5J&kl=E#JcoqnV#%fXz9GClL6RH{%jaPN6Gz@z3Nk~;8C7*#trkJ}ShK|jc z>&*pGYU!4k={$DZM;%(`p`jJuZ4Q%gDEf@&aD;_%@QWdVXod^FVTj5ll19HW!@)Q6 zRi0f-M|U|zx2lo`gW z*LLpzJWog3SKA4$c|LDqAk#RlNbaI13c_JJq@Nc-f+GE|FOddb7iwvRHq3unP$3gS z{Wd#2cf7ikHcGB&x8Jdmee7{L@|f_&XU*&~!zZg+-KsC4^<tPJ0*JJTiEn`1o#^Oy)747owUj-2cGBM$1Ec4#wo6TLHDQXl^{2Fnq zlRF||+|T*$;8WNjESjD81xALLS|&|Ic^ITJjD?)U0RfWgVb#-FSV?$XDW&548|l%l z;y&Sj&jNcT9(PNTd%YkT7mbE9RW<$c-s$3i^He6ZlODSrHQ7fDEUcuGUM@<(teZkQ zo|n&^`_ZxX6j+hYK*9+E7KwFJ`{bpjfao3oGI1E)kAK1IH$PP?QP*Z4YO%EojrOJ^MUE{~0?G zgN4-~bYS>s)+hc2v^*^E^*ZICr09Bdfb}t5bI=mPqO{~^D4A$YIl^W+fZ}9uX|kP4lYJ0MeWUF(QKogNebphEu{2U| z3s9-Jt}Zk=;a|06yIsHvH3kJq&!rxzej%H7ZhSP4?cP>EsD_%)PNPOWn<{h!%*led zSVi8VN%8Z=a{hRewuh0|I%m%9D^VikdD$kMdr9Rfx?6B!*#!*&3OJCdVOi<;`7AYiIV}mV3u^?8Cf~UF>tF@w zoFJ?`KYlc5v%Th!ttRqJ<443NxoGbAZD#}DgMwc8)Ag%eyLrQ?#=db9jY~US^_S`+ zi87iEq{tuJ;mzj4er^`RMXR4Z+5fmE5Q@hgOtYv9k3rHaeu+}1JXlz7;|y*;ivp~n zqO2A=xN7cvpz4fHgw-gVsb_igE`J7<|@aZAm1wsTW+#IGFH_Pq$Y?yB~cr$sL z`vO-@xH9=~{L4(7PJWURJsvR{RLXtu7jhz1!P(lR9jkUwN6ub!%1kE|9q?;Y7tteH zbxMt+E8tVG8Oc$|MQSxa`Of`qDCMx#cYn29m~=BUEm^!P&BMpVg5Hv)l@93H#&O+2 z*t4BQUB>M4xjaeQP6Bl3c#JE8Yo8K_j>O20xp1?jvvxAIa>OUo&;j|NOwQZgO#ZD* z{vQ66Z1gpf!0`@bfn`orZ?nEiZzsj2_KG7H0T(&%0Sb{ZSdl0CicJ#b1zkbX<3Ld! z54ieuu5jA*r9z|eR#*P5I>2+lq(w4tIKq2aAIr%$I{v1aflIN%$I?8RPge0-Zfh;k zJ)`xa5~+D??wgUdU(jI$f9NNkTg8nLimWjy>ho=!iJj2C2LA62p4b!aD#U|)F5P@4 zN&IJ?PFPl!T~wZ@;!Nn@^IKmL{`E=T8evgbu7M}b4OSb>ltoJ)fzJ1{Mhzf{FIR`Y z-#?9#hmERGqa?y#qvSt)lIPPn+s5Zxc#`v1x4W!EAs>cp^Fewg}RTbZUD8A z6#m#zuLF8bD{>cq5b9oYPi%c637qvofQ(T<#-6~|SYT^1G6fwFWK7(^EaV?RT>%M?Kvb`&qAnK-xpb15;6Gpso;exVCc}nKwy9XM`?!&#taMA>;#+M1%>{y zuLxM%hAwsudGLS#a<{hc@%INCpPdC?N(CN*>_R8-^=*lKmTAc$EUb}mUSTdv%#nRC zejut)6wQh1*=rEhvuS6|LMKgvE0S4(m^^}1_adqq!;tEL++wfnUB zlhnTMfkhvLMF+4nMX@w-NNR5%#YAy5{{(i!1$J*9(O;s&p6fZ?nC=k4ZMOHK0_(mR z1_x87E7e2>OrRbnfxwoA4din)lyfw=O>T-0Q(BE55%fAtsFL?4q|EP;)({W-nOOoq zWGB%Ytr;hCw^pyH!u7WIlCKfqtPwn)Rr{Y+gX;J1^Fe=CiPl);kC*$e`_(w~w);P{ zrt6JP#}Hjsu8g%kZ6EQ!M;!;UT`oh}0LW$NeTG%zt)TAD-JWnvcfm*o&|7Ahw;iZg z(aWj0v1bG^3V+Zbf6(Z9Z&$J3tvwoL8_>Fl%&LplYq*-$u;Zz!3x?~%hUaJs_Z}e?kMouF+m>zriVY zZaS*g_^a0JsdSQ08&4ITD>S9j6yuERORSVbX3(pEYgm{8o1rt#|EH}SHSEmExHiu6z#f%1R6#3X@$)fKyM7ob*eiWeUxSdV-b&FSoLd zV{=;5W^euGMYC%1$y{BFW2ay>GSdnk?{@x#>w77U{s5GkGpSFA5}!hVKr>iWeJ8#^ z4bec2KX@H*cpaO3S`mm_&WMA`K3s_YXH|}Y4(kha`Rj-DmqldwL!UZO0tL)*`IB|6Km5m@TI9cu3N&oRof(=5EL2OESuM8M~E7mH?dq5MAxVmApC(>MQ7LjO?|@K4{lIejWYN7&b! znri=su*Q)_AjkQA?HI1Fsycjb9u?6B_B5Hx>$J$6s_L#Uzrcyy&>9i z{4lW5RlZ=|VnLPDc%92(K;r?r?Hx_qwIgW(-;6X0qh~D_=2eTquKpc&@ z6@l1Kup4=}VjH!#6KQK^11ltQs-Aar_dVr<`*yPcT)6QRZod%I9Q$DR3|5&pzOrCv z*|3^f) zzR{GTL&vUm)m@9{C)VVdx#tGS(H~k$QJh6`JKq1Y zM6{gb!xCFev5(Iyu{CbxDaN`IlJ2jFMfJ5CD#+h_;k77?E&-HcEeNSTnZtUTZ%lBa z4|1XhTHGYlqi{v~iETXu??N3B>SoPS^wJ6R63(ENB^ z&wm=tz!+8vrW{Ph9M(%@tjEwzg%IOsuDg zM1r&s0gIf)v`wqm^5f5)b}jd(B8NNQAR;9y8CTKL+4jSY|&?2g&oFLY5PO9p3AyiC9qpCfafx~W7a=4UY0(8nkw|n_Btvmcib?f zD{E4^!e%n6mDjk~)={I56F~F}JDnVV+&&!wPdjT0iq{D`9-khprOr}RHQH=}VUdIP zlRma*J}Z^>yL|(F!Whk$_jt<_aw8Wt1{gW*>z&_}Hf)QpPO+cs&U4dGGqv09GYqr! zfvRV_vuBYZ_tEb1;uE@_`TwSr_aEiEe<$kwNBQpGiF*G*zGJ5TH&yRH>UaN6)%(9v z-v5>I{)1@vU#oQgSIYaJ$%X$T9SZX>L`A?-{qKvo?=%>1mZJZsvI$pffcn7>(AN;7Zb9Y$2aeP!9vOAa(_}1d) z&Gl?%&GCih(Dk-j@_VP`k<~4EJjWm5_Ud@8_NCZ=$-uxrbzsHbZ+~^*;{2nE5PtBX z5q-x(tPsWvH0tukN~EpD*PG+n*ZFwH%g5o+Y0m2omJ1uJTBYFn_Pf(fmdPicBloGy z*R%?oDy?0k-cJ~l>*lnbPW3YD@(FW{EOJSyzIdG2RVv}{+VWPHP^_r}atIioKJb2j ziDd<*duhR)w#P|Hoki>x(Ulsg*gGEJX10=9dEc$pPONSD=A(Es=7h>xt8A!1&^TB; z)KM0YM8redNbK)DFHQE0z-aIz&898NC_%M3igVYLxR0kymsp~=rc+rGL#Oz{gpcq8 z84tv4V6Wt0I@(o8F#`QEdq`wzZTg1bmKk{6N2Vmr=5D$n;CPeK{ElszFv-?N1!>%& zzAgICty}N$kD()97Q^Z){qm{|X3OCuUUxVV@kBy@fx&H=mmkh%p>@w}GKhFGbl0?Nii-MJ>T<1RBcs95$q~5{5`D zi~;k>VX3DbZJp4=Y4r!v^rFTB*=)5UrW=PWJPcK7&6vaD=-{UaY1IzKdf5zb+QD>T z)gtPH5(Ar{%ZT49};-&Vi~V3 zKZ7VUy{mKU{1aP-<_}iVne4O$@~i{x>#5BhzdCx%#ki(>Yh^LcY z#bvpDLC#w60sEF*!kl_NEv$wOQFFvuO1o=k`yO^x2>gcSZEY^Anj+EQkM^Qug%lfP zo4sL{o->;axkc38X#Gw`9f}r!+?6?UMb)w8x+H#iWmN2-K8?=(+33ait<^gj4au3= zugl+jPR@{HiG`nv;0Pz&sSK;h}hV$9nkp__SI)Fm{b#I?pMn z`NuLJy9zPo?x47((Ms0%W?r;1PjT^}I6|b{vG1dRVUgagSbHB2!@eVH1N0HZW91K4 zFN+2xxh0i2cQdvQf8>9pBd+~*eB#w2bk3Jd*3H`2~Vy+~O~9`vigPa<7s zM)MF!TdOH*k%RW)$ zM=CT;bY36ZkN!_P*9@zb+pQ%2xrf4e6+0X8aNGp0Ra^p47p zOH*@$PiA~WsW+9}Zz+=-%9|CO+6AdL*mX z+OAPt_Oni@c}k&oB_QZ0LPOVHF(K@q*`3~?IXV}it)-*iZ?4nUn>M1ORSy0X zf9^!mY2CiRZv(Vd>_pTG8?KF$nTtIFvhCH{Ki&Ccvs}eY#w1jj_J}o}xLJnUDaUW? zhPeNvrqPnX)NYxAN1wPk9ABf&(fcl?c|MdvR4gp_Aa|^$-QG1u)U^D@ryqZn*l)MQ zC$4RufZ6`N(-hNsF4}zMs$TsKY+(fWQ_`_`P1n81`ZLg>Ki1Gl|4klkVG-}I*7r-z zj?&T=VJYk~GYagUl=ZQpzJr5j!%Qyf>4+aiON@X>VLYJ^-C%4 z3Y#^|jRE#W5qc%g7;yErL=^4ZY0IMKLafBnEyvZW4>+4aaf`mGi`jp^a#p4$n>Aj2 zN(Z@>z?&JQ8{6M6%#8?+o#@etzZIBaMLV8y)t0u;eXgU-2HzPi3 zKjeAYUbja^k@$yvZ%wnWvA5@SN-2LX(lTVcXv%&wzv~X??Kq$%3F04l zoYF#C)Y{Qcb1zU+mXy3~z<_J4RGcKdSKn$O`b+@v7$A4g=i+91xaY4y`sH^B9>~^Z z58DIk^`h+S2OSdbRzEZe?+cb0$F`cqz^rmWR_lFhmtUx+SOuSWSnx_@VLWx_VO!K@ z0t}W5@!U%7+BrJexP6Yxo&DzQqIi2Ers(><&FO4piK+R%=k~pP-w$u4%>!3r7F%NW zEXooj!SeRYEXUw1he2sFxIFo|*ZSCHHHRNa7;rgQka^(aqYS{kAoMN+QE|wrZ7OGF$19kA8+Vk+%9F4CR9XO=-JiJ=6&E!~{0+X+rJ`2X(Nr%NKsA)vNwxu2?iUT|>k9-DgS-kIMUy!sstUa%V zK3t~7YUR{H5!MnHXPAU+MS6F2Q@(rO)X(LHr2-lm5d-Q1*xsQ_g<)=aE?NS)Dd7B+ z$?V0%{nP0ZQX~Rb1vCX1(G_E$g#nN%uDS9G>4(j(IH+ zuBG~lmsc|DTXM)Bd%pNh2QCAq`;{MX5Y3;N*@D>E-h$yd!{8gH1lfW_*$#1g?D1f6 z4Voj*R~0h1#tPUVclA;OKEzu-i^9358I-&gL;FW7KJQtCW-!j$6#nfl!o$D`+efhWsg!x=L3rM=~ll_515Wvz~{)ZDAnU9`il(r=7xf_|NE zIV~2zw_@P6+LBY>9(#RErvemG;FZ5t_BqP;1sIGJYPxP6d*saQFMTJf3J0AYce|Fr z5#DV!E(KyFyH9u+g#ZP@x~4j;GwbtZ6ShcV&9(?;BOvpUM}8e7untdpqZv+<*{gLF zxXnsiq6x2`zOSLOy;O-df1&oBu&T{Ff{GD>4IGPY(($ENJL&1JP&>wW7~ zz3W2KUuyaB$z)~IX&5{<%eet?Vpq0eQe-o;#26c~#@o;MOh70;6FIof8B%vKh0g?? z+>U)8rmz~mrIQ!(Lr6P6nrc*Me(|jc9UNgK@jgrk`-^`~Y{Ap`HGC?a=T5RBiHh_Y^b%PJXt(Q_6hM87du>8IFNoj zmVWzLf+17_U?>SwPLP?*9&mrljl;6-+c>L#{$`=gqHfynDoty#vdh$ArU&&MR=DaK zopMVI#WLMq{@^h7L^^9aRVg(7W;T5QG8N&{coJ05)h?Z^75BoQ0XViH@qMrf&7HSH z_WP0`AXa2qB_1-PsFj8_R6|R*wBwoDW>QjYqMu6y3@mEip-}XoYDt-z_A_V=?3(sV zYqdJ;`CnF?1oGa(cMAHE@Q-@HWMShgi*_WO26=Qn#QJeWvOr(x-mH$G_fKzWrP@J~ z?eZJ?8N~CZwFYk$h%OAh5)?!bhMEZB#4|HWsdQ4QkH-|UvVBck)L&mjpjRZ*>mkC% z=lCE1cS96+Lr+7BSVM}!>5p5nk6S_)E(72-cvF#09D;3MRo7QFujwNOVrWA)r*M_9 zaFx2S;qJTI=y&zZgVT?k-)_VrZo zsP=MgNt-Fu%ij|x>ST$wP#fOu81=a*bs7u4h4>cAfS7Ux~TWg5vv znh&{f!&GRrL@nLIswDGwF3+zl*jYv03-}CJq}6xT?+IThItyHwHO|G~^YiaZVaMWk*BHoAV9}{0sX)I6aE`s zv#tA^PC`Nu;n44FR?XVUpJ%#C{(RIisNg+yoOGQv+xHkAg4eVDy7Fv{@d$kxt!g(z zYM3Chp{H{Q);WYMMhSjU>SD@=I@E_c#&-_JIoe0#Wa_wT=;M$Se)` zu|j3DVz$sxSsm)?k?Kc}nVEyIed+daqY>E~#k}?x6y;ZvXR#y~u`4;DTd|>A1)dNr zPe@mq_be6cMVjMEhd(_{D1Fhu1Wi)@gfteDpirzm%Aj=rM~LioB2q!ceic!Un^BqQyG;; zdaLDhZ(81G9g2%Vlr!%tg8dT<`7YH_VJW29-^AnN#N$7y=G-pi6yaU8N(cKp7D2~= z0}T@-Tc}M!6i8M+>QjEc+db_XBOK?1fkl;>!cA;D7Uew={)U5-fghD>fFUOJ{>z_+ zF>S-=n=CC=wxZH+k1o%v8t!Rz9`>CPu7U3io-UHI%qodES$LS97z!9=;Roc88Fx5d z65Eokabv6SN4+B`K-vtcFZkl3&B{^Z^9&$mhI`1i84w562qR5z$RE1nq8Ax%CMYU( zU9?aj`Vo4qqS%6~xUt>%F`|ZO5W@&xlu?<4%*#9Ib+oP`>Flb6^jV73clTP(J`$x2 zj;odNH{b|cwdiDOh*bn`VzGZ>@mUEMVj)+*rHqQGjEbX!w6h8(id%%9(k>*<|FQT} zaBwC7hnEpZ$-@E?#m{9jc-&vvrxdkTah)~)M=Naxys(u_GjmDfS3ln54@V!UBVA!n zG`d%MKhWB(F3LU;JcW#zef0KGo%mAhhD3T>^bvMQp5pO0A&Na3JEeWVR`F|jdo+lB zgm9~8(mQptn_F*XcuxHR5N*k*@W)7R5;%ZDaVk_Wql0I$sw~;w8ARNE&raS-MFqNG za0lf`%@Cq^qM==^OebFit*U?Wt2StBlx8E^J=`o4yYh7(gX+aS5@q~%b%!k?|=ER-KHm>=jz zNmUU|Rk<=6k1(uI)c;z6St#&9lIRB9j*a!7$Rp;8#ekt~P#}(PJzbfn!jb`k0%QWSB^-EQOy()%jr0RTwS4 zlb6vHw4(mDa({e z!x$?J9Bf5OKGRkhbBgmR(Q#FrKN?2SG~7K(^|$2{lGgB-1_f z$zT%*Wjg{qt?Xh0u#R@V-HSgbzlv@hIw>G8dkQR}b~aQDsbD@#)3cUwvi_(;H1hpqoDSOR^Kuu4f`wV~pJYl7Az>^b zAwJ6yu^iY|6QkCsB_AiQKvPs(Q!e_glBX_{va$QFx$oc~X|X?=%4g@k+LuBNvK_1` z(qed8*?-Ab<})r)=`Ymr&Gunmcr?xQM=JzRRux;C&KOtgMx9&)HkwbY-_0pUvO)2j zOOnY*$)ShIfRdwdLURRRtq*+LN#Lj65``vkuZ zyId^RAsK(&CkL(b>F%*VXM8mT?qWRN?tekAH_Y#k?x4_XTCXO{fr&9i0Y#Ij)dP}l z`P{zdc}(!929h`db+In3`;zD+9EapkK$et=VwZWbj{zHc*WNP2@J?o}@EU%W+X46u zh4ki?f;(UAjRy7wz=c}wkOn)E*Czgnnfk_y6HG!0+DYj`ZLe>4@EaLz`mIV8arndc|_OcFPQ zmsB&7yC#CUvlu#?K^~6-cqLO!&&fc9WM5pH@WdwK1GC>c9DnU$rg&!d{?m}5D=MKC z@f(TmICAFZ?xyuk;nfi|r#2TBy)e7z)%D}0l<$81+{j-izS$0;lIVe!`>cx{=(SM* zenzyH{eo@s-7s|)WXIh^_O0GxxU$H$X^W+9HiA}C!REM+OgkpvoqIJSuZ{7coB?*J zc*a9fiiQG)(#JcR4|ns;ZL)} z1s*jb95v$kJs$$m@_DL21_?*VX{LY*wu#D%gH8U?nRrw+a7K zI;9&+p&RBuY1cW8&yx3cSwF+7@Op1lv=Lw7RdM0!sfmoK&47kk#G$k+b;}i9D-Crk z4cIxQ3w;e|JJ~nUyRK{XbW(dY(tA8#*DPly?Y9;_DuX>SnaAe76Il9qDeL+VxCuRf zZ5kE0_M6Fs^R{XA-R-NvyI8CL-Kj$bHtTk^n#1Fu##KM_6KsZ$%DoOi_*O_c2_^ zQx}qB7m^i2|4|e=ZkBTiEPAYPV#qg~&hG>JQEyZ>^v0G&07S_5Oj*@o9N{b3Mm zTYCaqnLzugn)&<kV}O4q;@`nm$YG*yj*_U>cSL)B&*y_X0dA%fMSqPRp#Elhe*e#%TQ0hw9%*L zM*OW3wE@Mqf8Pf#yK>+_ZFb}Orn)6(y)^sVu_e~ospw+^HK*R{EmfIRKmGB957Vke zXjhL1;z78xyvtfQAnhv2+&Q%Txnc4ub(|{oBSYM{K+C3sf;lKUnMZEmE#5dXZJfNq zx(k&unIERJ2)@|PqrM0um6jov@KnX5$3#b zi`t4uqueZy33DjvYnhk{aVl2ZJmhgGo4BS-+QoI&XM%(?(aAjej*)gP2=5d{QzoRj z2yi2!I?ds3e#z>7jV7r@xm{Gy_XXCw*WJ6!dp?b(XXG!yI9mC6&MzHEXjv;+!Qbkg zLt)hgv+CZtY@j)BMEvUQEXj>U*LZ6DSuBv+mpM_l852m|Dj+N^Op6uxy46eIbt|+u z_YS_;meVm`ChcClMi^23Fh*F!*s7HE{f}INkPXco2Ak-pZ1QV z?R$|uUEGw`9=n-LW4SfjwJwp9gix|7o^mhM>KPBI>vj)yMp*o2FS|br_{`)W^6Lpf zaw_x*&*qIM)Bg5_OsW%~1ojIZm(s|SQQ1BqUA;10XTNBL0Ds#UxF3cz{%vDGIz5N% z22mR3j|viOyV#u~*qWL{A=3ks>5;MueGL?BJGk7du;u=SJxC(x*{O3nXvql4WNJE< znEH(uo8K6OFOq&QO09bw8hksui(TXAwXM4@2gcHMBl&8@e`cGGpAh|UqP~4!Y5p$!|KhOKkprVMHcTxq%3?IPIfn%m&`#1!URL3VS@ zb3>6+)!bm~&54S(Id_1oY|b?JvEqsEV9o#KKeY*fo(6!e=K^N4Lb{g`7+^>xR+Uk5 z?3IOHF*ak>T}O&6BkoNTiqE_PYC=58cEoMV@K7(as!vPY)~i5#hcMKoYFuK-Ah9Y+ zfYcwMNq_`g1p-&y4>ef41;OHiQ8|JZ_R74`EmSMUDt#76{iI_O#j0BV zF^9m$wfMzRbwnUY+|*8U3gsgmr(Vm`79!rfwFI6mBCgFgG}Z;&tKBb=y20ZI(ya37 zu}18%CJ?Ds7j@P?rY#{CV*;^N-M|-nYs1cq`b`pX{OuV~VuM6sJqMn(M5KOg6t7As zp>|vT>569;`5aXkZ62t(>kztgn&8R@8B_^$Eos4n^s59qmh`HjbF7q7ZE%r@5Ci?ye z=6F6hr#U!w0_(l-LuWF6uR>F0oZm_*tq@-;`*-K`)k|or^J+rV%YLg%2bIaWFuDD7 zA(r4w_q_?L(AXKQiANgR;9PfzS)ThFX^QU+eGTlx#twdWm1P?nNH=`piYQtKXSN3C z?E4WHHg9S4R>;ea4v>qaH?WHDEz=Pgpqb01m;G+`vG>b9aaFcMNmXLg?;?+e!>7?KMdc8EXjbG1C|I*820A=06v2Cg7# z(tGh~ir_+V%9qB|9S;{IhO&SKfK?OWK^R zWyj(>fow6bJ3mQWu|V*d^B(rQP#k@%J6IaD=@FwG{Vv5%tkcWe2lj;*qHO$I)x$2| zGes%3Qu1rC{jW-%99-Nxlq=3L(C)2f5n+TRC%_a<_kNvbRdEY z{V8Us7Y$wvDGFGDKamQK0);X3IoBBl3~ZnI|E=3f!mc2MxXm*ZNF4w|Al^LNXZucJ z8YUh=QLI{8cBW78$%6<11wB+q^HV`oN9c*DFrRAhN#l(@llK4-$>76h>)>Cazj#%j z1K~Z_V}JQ}A|k_m%k#4A2`F^HfDyXNyTjW3TusI98cbt|EdNsW2v#; z$A5~sL?bIl0RLvARxJa}^N!(`YDJ1Q!M%J}ip->9-F!zdCLV~v!A_sLN2e(yqQnip z^96@F{gjHrZmA#>TjKLUSmFzAbb?I%2v%4@KL3iXK1doGgfM;^D@eu&YqiC;q#iDn zjD$iO`fQaUK1dS!JnDp!_L~h5$mI~-<2w=`Sgzny{dHlxUUHarYyZ<_;U|jon2{?< zT40Cs*B}kybUCmXe<6yQSaJH+#7Q~K)K=^u%yFJTtRmKLyE*pHw%NBP*67sH;xC!w z6bR9Mxcph-5Ry;?ry~~DSRpLm#osvf19>jWyW*z~;`e+$0-hP-VHBYN{%1LqK(msr zIq@GRl<-==YfZjXkm%0)hti8S@*U;%%kMxl^{p0@`iYMi7w<>8M>JH4u_|m|snPtA zosqUk^1pKAqkQjza=tMs0xkO+vALn0YH*u|wNS-^?h$x5GdrpA}9_StUi2Uw>SB_CtiYbTg zvJ5pPeWhru+9PRg4FTqaR=SKhE`vYtr_=Yx5_#OMXsW>XH&LOKVWd){F7x-~2a6`^d-sG%vmW;p*t(H=G_@+}2P)nIyc&mqE(_%n)_z51A z+WO&fI=)Eyw*a^Z4phR%v0@iI0=i4TPy$rG z<1ajuMOt){D*g?VzozO$8y9Zi?C-TA$$9c#w!dBn=Es0IKZO!^Vy%~>zm&Yhn2y(I zMj%eO-!b{|?&Y<0tG-_9oWzqC@N|8Z4;i`k`QHLA7me!}vWDMXQ(c1%);5wJ4~YyG zCFOXl#(9PgpMGkbPq11f$KdAZLeqqvp{m)$oH=Hx|(8tq~=E|bZXc)xfMES!xpH|(<$}Q!xtzwn?C({Lv;^~ z{)gl%&VS`${paATf06V5TkzGt$oc;qeD(jvVg2XetABy>BzV49;n1jmi5&6#35u!R z*gY4>6+iX8Zr(tF>BX&qj@lXuk}dP31nvAvv0TQ_ZCd%#SoksczsY)|Gco%> z3pf;I-D2UecQNIQSVhQDSri1h3rhp{u` z=j!m_PI#t|>_ApEaNA6#QI$e2dI{Pokvu#NPwO6M7kum9{-rAH|Do-jgDeTU=)vx3 z%(TsE+cu|d+qUhlc2C>3r|oH*)3$Bf-2T4(VRs`oV*lBQceC!x%(_`o5t;Q~<+XwdNI25J36%r6x zPq1mAPpcX`#UDL<-$`$6d%a=$*+YF&4?|mz^qNnWr3%JoXf}6{+UOT0kCXPtKW)_7 zLBSM|yX;eWs5ELqx+6ZVIrGPOYGW+_C3%l*Ll^jdbV|t0J@-3cj;`pA;G~Tb{^TBHM3K##>c z%-J>BVmE{%??huaJ9>hAf^R%{T=ul^vUXQRig1Xi!pr4BY5UUT`)}`tlsl+*O2P zq=I<`kFB=|Kk)9BMaGPaua>#wj!LWWg;-|o0z2{-_(+!D%95ji4bY0ixw2F%AQR1n z+ENP^WIcycxVm@yAqHeFDiYAR?9sUHl0e<4#bnRf8iht5?Q~L}^H||qTYt}okF4;n zsqOh)!Q;Cg7^NdOr%+u3iBb2UimN^FVqX>*q5i0RNIA#7v}7R4>)S7ZP8afS?EIso zbyY4@TdB|5`kc5g8wfLCBLJOy2`ma;20@hv+V>zz@^5okw}qWUXE8EECY)A+5~g@> zu7>`dM1O6|CUT$|^zPPE9}8v@ z>z=7-eR&yqET08F7tJNdi>q!<%jh1d*iH7!P7+116GxpbfeYWORW91P2JelHar)LS zAG;ZDZ&}-vTV?iE?TO28ZG{!@)~!9QvQbNwOkmi{Hyd3GyW5MN4$s1JFC%9wV{2nv zch!AEa*wl!h@TMVYDnQ>JRs+mtIN4Mq5rR|-v3_ofd3oy-~Z36o)aq^2OFp5IBq;n zvU}bpjV-tcKN)Oey++uV{dEE81c2pxPjgMwx!4>&y~i>bw+nn?^l+B#SuxR1yxHQC zkYL-h?*-BCi(9=r_4W1t=E7hycn~pEuhn1b79T&V^)j8Mlb-vKaJZp<+4fa^y+AeT zk{b)Ww@#Ali-<7${CKTSxLA0Q{=9JOX-g6qIoqqCg=wHuI+^AFY11~kI0^7!R}tt` z+)Yl`|3w$8-<+(rEF1H-M<1(4Kxe03h(NiGhaW>1&{%GlIE^`3IyX`@J}{vE;w;Jg zG<2Tz)&^K*M#gLhDs4^%+s}N)jqPcjy^b9bpk*huWJ(&2P4jIZ4@T}QJG<{`%T521 zA0U4rUK$_uz3)#qJKHrXlYxD9J%D$#WdC-ekMw^qY;Stm+Ba;voO_$lhy2UUI4L*B zxIMRlwr`Cj&o*{|yfIgan*?NS&;os{^vA_bP)Hkv2o71Eo^=~)|4OH9 zV(?l6z^9FG3*F_<(p5b#wb0jtmp7YFC>(C?16cUrz8(PP@9dZzur#)w%tfWQ{?6X^ zcCEr5_=^_bWd?nn+2Zw|EYes-=u5ZUGdZuql#&gz|4Yr(XDeg%UXMu}Kvlk9YMH$% zQ#fXPhT)W;m*FJ9pWb-8GcmFBU%J>}N+33F>_Yly`JND~ix!KIJ01(?{W;Ce{Ak@r zdUkc9=h$jtnkin&a-Xy(*&n!xhSa6Z3>5^}!NiJU_6HsAf3@Us#V?@acBjk7Tw4D1 zvR#Xnji;plv_m8O-*4|6v*q~bfJt<70Udr|z7WBtrY;p7t)u$eWm!%dmdBny`ZyijIT+IB+C-PC)$ z-&%Mi9KC+Zbl)$IeM-H54&0mE>@Q2me_nEao@}0wg&QYVj-Gtw!tQJW)kX~>tINfq zjT*Nf`(2;r3*-TNiU?U7+O4h9fWQsow4{X`h+oF*+oPV|iP_lFF|TP?7tLKnN(V$f z{e+#O4^kA`dpqO{VF%?5^*SZbPu;!pJElqgM7`7Rmxf7dO54@5lcO!Afo<>A5ibYn z{~RyzSovS~-LyqE=lNgTPQ2bj^Iu5+b$yRA;D-8Zwm)|FR|E8qd9zFDA)sZNwj<<; zp`8{Mn+9=Qo`+B5*nS^My2{G{U>u3trT=ss)NGOxcL~?vUix?aVm6yaeVb+MP;!}q zmnspA+=MX`nUmq%*&xKFaVlUS3zjaGtHlp>N2|V>UYh$Q*V^o$SS^fk%$~SDyG3s| zZeKq+{@DfyxY4?d!S-i=&A;BYwb_>gOQVtlGhl2+?t#vimB9$y@Dvob%9*8|4ySTn zZ^h(%ZSA(Mb3et+mDVgh@tBT#(F*qupaUG=@h^;W`(Vhfo_1LjHBqQ$Vs)91 zUuV1indhkxOu^~&jyP?<-r6z-Y7M(Y{bZ;PJLjcKWgoK*u|zD%SCy8jqANwZ5gL=U zbr$bVi5v1ve7rJH8iU=m$~-$?r|#u;{S^>HAsO#h9)I<^3aHd}>*V_Epx^vcvIbAr~ zDi>?M_)NirZwm96E623IO5iJe!5a(ghv#}vK!@{Pads*V?&h$%6=u&2g)&m|%H@?> zS7FLfnY_KJ8vddiDp5<>HYPG7N>n24vv;Iac1I)46uYKLyiN;j{^*jG3)Itndpg-a zRpak~od5veTcE?|J`{iBv|LsUZNIXUd{663Sm9H$!+Yb7>*@0y0P`IjDE%Nld!AQ| zIKJHq?wo#5BD?n~^(pL>?85JNawYB69_spv`PP=R;dA(Sw?VeNbJ`kQeMXypT%I^S z3g>jW(f;9A11?aluZHT^r8zJ(_`jE+#2Er3k~*)C67oEJM%S zkk&lASNiknrR8r1RgT<0It0D;%mWx3t0FC(FUbK*Eqsj)9W65_#>EreyH=_b1ZSOh zK^`-+$+R!hi=V}_D}wu8cdK=umoPr_jDZt0*C%(KC7&s~j2*?i*0%@lde1B_= z@7$#A7GD+$_*;LO&6nAJncbJ!Hv#?*s~SjOUA0qXOFVAqK|pdyK|%OZkc4@$b7Zp` z!LUHM;P8X@HYrW~df%crF!&^3IZIn*F6(4P+F|toeQWgx%Tl@M zpoY_20lvk+%cG`cTq9pCvM54XzZOj5640*7%6+!n4#B~@v@%p`SMOZTATv8~>{Fo; z;^HjzJgLXm78>TT&>(uAwpxZUKz+cfr*5y|P$RzpNUcr?jK=BK?5XX`uH`YTdutqe5T8xMo&&-^_mgYv7@ zR%m-rJNId9YC>?-?~?WI!+fylLB1;h3t`(m4Z8%0j~fV!-Qk^Jdi2M5cKewHcvdsL z2cbMObBF#@kOMzF19Hj#Q#46Tt&$wsAm6b;d1`@o&<1Rv3)n^EHpxt_k{#Kg-2H{} zu^il@d zLlb58|A0b9CMA)k$h3#1$P|Pw$?c`Au!p9}6oe@Owt6Ys?V%Ym1z}4F`Y6Zjp;Q=*T{!53ZFq&2uPpvH)GSHe12VASYfn1Tr!=0<1Y$;y73&Uavd>2iX zHPyInjd3pZFDKz!zV?V_z(g^y;-CCJOoC1R>u4_1zuAtUtx6^PqWStqFM9w*e>VBM zb??mzT9Gm{%}nONlWk556GjFf-Pw+smo52ZS%n?Iduaat9u@AzMY~@!n0$2C5x@9v zxbUClI-NAR)QmXkzi+il=`Hbf4mouWlFMb*S5!BLe7w6s*-6h^Nn*Koktv=b&Zk%h z5K!~RKKs{{aJxYv+iv$F{H|C>yPlRNvZwjWQLQv1S_vLs0!anPD7a}$ACf3MTnQEx zmk0Mi?WhCS*)Za%)InWBOE?^BN+^3PKSw-c>iXMb@&S96iHB|sv)bO1K*+K;9bNT2 zt)ZoE#wF1^r4q)0t8F>8(S$L}haFnRvVobr&1l5LP2z&K0hdfOcwDj6au_kneO-{Y zB9Mb@d8h(4t;n1A&hQ0BsMd*FDV6c1lOq6Bb)wF6DWt>jOs>Zrap^bkX*ZwNB5$D0 zK*YbF<_?vp^Ry}-js8qsIlt+L5ypVh}4!IKWqOp;D$p5;%d+qdZfSLuKm z+_4g$QU!7bD^AKyl3|+p*%vCM#pW0+&tW!nQ06ncbjKa-e*Lb;F(H5y{x0vC1C@u} z=jzRR`m7isFC*s1`=Z#9QNu0f`B-WYZH~`+{+Si?XzPA`g;tPBCGFIRk-Vyl)qK#L zB;#cAWWH@>y#I?-z=c1m=W5B+Dk~5?QB%`tFk`h^GQP@c^B0U|#NFl<)x*nnkUB@{ z{?&&X0&qVaERfyQyoL(hd-LWcoC>{6IqzA!b1{^e^qRdW+0q;HEpi8*&KQ;E zGBHBeOe5XjsH&_g<>Y;3Eo703v)&F;)4;HWX@S9A^-Xp8rX!{sLgYjd!38h});(JY zghRm+N?BCYO?jK7_{NGpL~u{RmrjkPU=GIJ!Dp#q4yu`0RHg3U&=+_lZ)y`B3s4Up zM{yoq?RFb>nPs_|2TOY(y@$-wT#IRy0xNo>C+W^?H&NZ@isH~q<&JSIr;&3B*cpagMgp=2S`Qv9bVG4r zgMCH8970=Ktevb5Q_Y`fonVCC!%w!GW>625Qd0!jE01aZKE1Fkzi%FVMD;h#E>EBA zIPii$-^JeeK|b)y7KXK!2Xhj_?L6!-vFz|T03m#&>4e#Yg&tg zDOA>8EPIcO=D@SI3c%0a-7n;&Y>KyHY5-}lVO}VymgKw%�lW&xu|_eE=tF$)<-o zw~D2PV)~fDP`a&bu-+wj66$p->8_%V9MbpjUe443LDpWz;PLu!)>)Lc!G%J1aDl0R zG^vnwOv}El-0{y@_^Jhlu)rSNbtJNr{Q)=Q_Gk8PoA1b?2QY5uWgH@eLM!qhx&9n3 zgO70g_4y<^3T}%ifdP z{=TvyczW><#w}}^v~x@Ant&v`m;YGQk{B#ktiu{qy%FDlX3|ZH&00m`TQdTg+4Zy< zT8murF%^CR)XNOd6cRo;)6XiGw^gR$`@wPu>}MdLSia64Jqx9LJiuFO;4sUc@J^@m)(pi*$Q49ORM^C-80MwzG0S`ea)Z6XMzSBX>l>FXn@ zEhY-wdrEP2Qwn@C!;H2xjfGw+ST5R+?$h3Fdj)l}uG)D>`8>U8=2&!SI^^urK5Q#D z#tcF7cFN>gza08z^x~6@XiQ&ZE?G6Jo;efJ@tZLD*8CpPa_xVl9Gz>R>Cy7;e}tNy zYq;vw3h3`(6=(jq&}Al}om0FQ2jE?hHY@>L1(j2hW&7W5iyC-8JNCP^iiyjZvPNGh zWOQA&Drb;3sItNse|8*pYh@EBQ%$xoEmCnUxc{8Zcf4228$mA-&@JWASvwilnTq{4 zwO@iK+18%KoL`oi<|+0mIrF$M^1Dp+U}T!#Rx*2z!p=pSFXM(PJ}IAQn>H9J<}wv{ zu@V+R2p0a#uns~9lewko{Z#z9Enw2e5^>eF5#rqir`zq4GFX>Ux3hoyoCJ`fl8Mq) zqfgLN#LZsmW@#X=jiOS%%WnBn_}pc8N60ARujVjCJSCA_@CIRIBocj?%@T^k-Qo4E zqCY#It$QfZ8Qot?6y?kA4kF9Ju@kID8;=?>z`Wk@iNQ zK<%zdl?oWsrAO}4^^r;el6ck)9+)Cm!wT7qS37w!=f*2|`E^orObXzZjUPR%@ z@dS}KqV>KA5!wm(!wIZ*+)g?`zenEAq^mkW6Y){>pEHHn?MvoXWeX5vGB=dMc4Xdp zCRQ8Y1J;=-;4T+8Kxbi7jABW>J?j-t;|f)8B**bJpGjuk+!5kV4~F(9tzc?jwhFQG z;lp@W*C#!{-IegWqHh?PeeI%AZtz7uu8hvh1_wXxX`Cr=c~%>ULsSCrIn|KQE{HA3 z2aj;%?xA+pR_*6M73wKK^pXNNhf-kB+@jl1I*kMH$DUA&sj!yXC^dBC=WWhPC`auO zWHEg$>LeslI&4q6bamR|J#qyxiMpwOdl|9Sk)idyv#28V-#BUXl_$??wM%-FxLB7@ zp4q#<*LIlucq(JbnbrU&k8tH}3cuN93-pf!Mj{#{T1c_GQsTZb1`Y6DKCtpXsM=y~ zD;2jbZa1gX>VE2urKY(DWM%*kG3`Rt(o2=S8Pq6?^M&H#A!B>BzdPDEnS@du-;<3B zLuWwtEal;2#Tg;c=X}jJszPB=-0C|}Lfar#kLSR7as7`V-IV|`mKr7@JCobs4kCL8 zm;fXMX58+zsJ5ey)G3z0++h?d;@eGws`93}zQRG|27?1g3M&_w0vXAgX36IE&siPZ zEQ2mx^wo%>b(VEdHIIFUU&gFP7qp9rqls2;F5jRokK2~W;hfP~>oPt!HyUZ1qDYn> zKfzF>*wk9J_AxMc>VHj{s5K5`q-QUmQVX1OU_^Bm6 zD{{y5%VQP2yrAgnhFxc##WqkWBdwWZBO_FMrRzRTlbE+}f&PdK967vU4!T^^b2`P4 zHuM|xbCH-QNlc6pM0FFH<0|M{rDtlMP3A*&jwEfBsic5a@Jkt*N{gX8vcR+C0k@DU zszxkE{8^jmjP3Lu{Cb+xrT@4P%CB!zXM!oag@UH}6LZjm?dT=xuEoh@8R@=Mi5U&x zaF@Mqct(``cEL&o9=WYQ1bQ-7N6!A(c&-X%vD&R^#C+=Ie-uk8`)I)}mi1fPBVGT- zm~Z?Px0M8<&3~0*poSGhi?SBoT(@8K6t8GwA^gk zl|5-5SoJvQMFhQb?N23Fw>9Zcc3Z>>cKW0&boKgFiI2=)2+w!sl=#FI@OS07S$T{! zNsjn7ge1N*dvdlM8^Z>B`Hu0v3(@c{f7@+c8^#4%Zyw1Bbm-BW5p-1u#@(|0)<+eM zq0N32{Qfy$K@5+&$c5EnLOaCumz2ebxh`ylOX?yHc1RjemRmRgbDJot37)sIJZ zYu=>3W~RYPbrJl;2Wv2Z_dD!Tx^0J;g`xCyzr-8@E>@^HJlE$Nf_Vg?dnovM*l~rv5;5x%q2j8#s^MaGpQYk1yGtv(Pb+&!D|<{UdrB*NPAhv!D|<~V zdrK>OPb>RKEBj0<`${YOPAmIKI{~7e2~op=qVK@gy%Xs-paB}e42Ev@on``2J(IMC zLrLG^?^lEWfCg9u^LKPJ1R9(P0`*Lh8jggoQ+jqH0|qo8BA8*&%`j*t7}Qf>pB(Vv z>N`PBVQV%(PT^}lKvM9Z97s|&A01#)aGo6SfV~0#I22z8aXx@@P=GL+q)OywBmyMT z+d=yQ?)qFKuUiX$q1AWai}2|leRN>I(whj*dG_X_K+{AQht7<4K;3FUos{uXU!>Ka zz^js=jpV2xa>tgSyh~M;KdDin8g2m^v+&!=YdKl_AmLE}R4eO_rFMm9Vg*^+lNDq! zAcLZgW@7I}h852?s-u$FZIfz@$n@HuPZ1WO=7gU$k5g&>I4+`amn+L6AO(e}#T}T= zN#@dAn+?x?|D2iw&3-SDnuE=LUz+S<-{+$=qsjFhmpV|EOsE3IFf!ZqY$GGD1g+;$ z^k_Qw$(QJ^f#{yDpk)2KC>+`rh;hjmv8JK0Z1mD#inK-6N}Qq{WWK8N+)sb;z~7#I z)wSV7&QK+zN0h$B)0#A)X};lDRN57k>zEYhOD^nSlGLv2QGWS7l~&M+sywYp5>)S! zC5__nB!SkZno2c{d=0~vaMyVQYF0rAa3h*Z-VH>5Xm<>0h{YV z`RROk)}U~7|365|Eb78Cc{AY*iGh;yRu;6`LSqg{jFGc-qknJT7iH zuhP*G!|A?eevVfaiJ{XDPt@C(On}Cmahx`ljy3talPU@>5wBDGQS4FL4j>XLFo{TS zO5?HELhra5VKN;bY?DA+5Hd5~li&AZ43Ca`K(17C6w8klP=vdlKxZfQN>$VSZ@@O; zrs?PY;*JK>P&E8dJf*Aik7U@|*=XTXTtMzMyynNqx3)CYTGdF-Xy<0-F^58-62ov} zTtPG;oW=r44?)9(;+Gox5kRn;(?C!HWpzj4pJj1hfgwCK?l&|g^<#z-%-%BGPYLSi z>W6#%?ZkmlB|n-0(SK>tU7&g)rC#`tx11b zhw1W_N6MA~PN>Huuk0dmvaO%wf(mUUpN$kR(=Hb~BF&NPeF`?^rN7UKCCH=N zIOCmfnlgwa$emtM2upRcA6^(-IkX3)c-&gC5-o{WT|6%g+RaaN>a!!6@J^^DP%V9Z-fM44`-Z0WxSWoXl>-EI(t~<`~J#L4`17UZ|ET5L-)k;&CssN#5HjJ{p%6>jfD*Y z;lR};uxR3m?NRAx(4WG!L!uvG>@hLKy<)sS_5fPU_+U@9?_dLI$;agpX*wA5vg=1{~4eUtL7BPBxSr$17*AkJ2y1YdW zgPO(}G-rnCv$_$xSkYY7Z(C3Ry&|5zCV{*A*x^S>hU?+Y^xncIwhQJTrOsGMrAu5AL@OrT$zeG(TFH9PY1O5z@dR<9-bp1z)u{q^hJ-_g~LRxIj?b zT9q@nGVa!Wp0q%FJpL%j^2worwzMlh5%48InInpQcqAcel}zFqLN{nLsl5TvP=T$>!cTl_u)<_KSc8B z8&)K+5WhuzSL&eaARD)h4Z4-1rzE)S>VP&(sBQ)ci>Giu2+w z6trmwVkk6%aLP6FBrdKXo)#I&hR|!Jgp$iNGj}aNxWEz$bOE`|qF3V@mpO-;a+a5%`-7H$JYCv)8ot@qHb z9yk@Vqv{dawD`*E=7y}N?sD#VJR+9PWT>gnvx*IdQX@?5;vrq?E)%uw1=N^?pr%^6 zEvm!W&Opkd&IBfi+2|a}`6_I4y=$U-DSu2#VS6Qn!JG!gWhn9}oHEjKIe_}#^u~CJ zb@1z}MKZj$US0^P{eq#A*{S?P?0rR58=>zSqPoqP$GaCq#G@K{#7CMeCdCRLhXZ*E zIoX{*CPbDa+-)TAf_E?D54;Gi3X1_@fLPOhE!2SbwpK<#gOr}9R8;r=p75oPXg<+HoQ)wu#KxdG6WCL2o*t* z9?;oMgVXnmj-m(ZgeDaXj=orQ$tu6NhcVU$SEM{nOoW0i&3Obt2?$V7XA-5WJ1iVq z2Z@KSLPm2V+T%?K*Fn_F>?ZLskj!J&6)F6%wCq7g`vZ3$dqbvWVP&{4*03w-Q?!bf zuR7Sz&64FpL2s13OFa|{Ne5l_2Hq}*gL);z_y?&vL`pe{MhiUdo&eeBF;Ct<2erRcsO;cd*fFEcE zF`D?5Rh78h%5hiK?7|KLOeqh%F;zBi8OP5EDqNzF#XYravV=tZquAG!nIpcY z)%o-g(?b6!U6ze)GSyI_vqY{YL(pwQ`Hn;N%;~(|_ZBTmDxVY~&;%mLSL27>8_A1Z zvB-c1HppKQ0Nj@KQ1K%uO*6r!Vw+u;cXaqTc#wX~qNNz(L=O*~$gb~GZw}yh0qt0q z=dB2PkwOKwMZ{zscdIB(mzvQo3&;0lu4Y>gAzAeW+YX2Q7jR@AwQI3u>^7gT4y z;iIVyNnrGiMWRZ&0X+sT;Kb;0qhQI&aL`n3ohZ_36LaS@cm^v7t`XH0d9h`8eV09R z`6@^aKVZX7?Jm0SJ~6BE_p&v|1^Fd%lIWDCccb2dj1mio5oqMWhIx!wu-TOI1?60M;o*203!fAiJr1I^r9J`z%`1P5+``ov#=_m^wC1 zMz@X$Pv}d9(3<>M7<9J*%QP1}PA(piObbT#+DC@qklo!s)4q#I^`e^K`sVH58~7Ug z1}J|L*m5{?bC7 z?XCR$mxZKt3c2&5wGix#UueiH-1<1n3&^l5uN+fJNh|CMjPhIkom^Nu@X2C@c@`c zRhE4Z|ETv#g`K@@a{@L0<6^zvtlEzEBktT^g2%pKF=wN*=ke$u!egO&hmQ`R}2TF921;q)vVA?{Lr-BNg_ z8Hc3Xca_7(t==y73DwnTcg8qND{yNbQjsz` z({(VAmcmX7(cMXdJHHm_N;5QR#`Z&sxYL@IQ$JHza5Tun-3D~3-h=IfcH3&g3%j@r z4^GT@_7b&jn1&QO`FlS#It~SOunbXCdDEu2&%3Z;XW)RlQmzpLl~61^fa`h;l*7*n zrg|GV8W*0;t6$_sJo(h|`t&#+0~wFzRvY-Gp(sR(eu+@MAy1EdYvXDT^GJBnaZLrs z$n_X~C>21~LNQ8kHTYAs#=MgQN0rEi8>8+xLBhkISP7`$MuPW+S%ItxEA~+BSCw3~yI* zqi485yQA7rn6YA8zc%N5(^U$^4Nrp;PREPyic5e;f;@J21!KXz@Ui0L&eE&$+$KgjFQq*)LoSz3@C8Eq^qY)p1mHr=yS0vobaxFW>+DOIzMjB^%kX-Mc^ z@i1k62T3$9;Q`Q)e3;E4K60RcC=`7o#ezx~O9!8UlJ2VLv>>lM^kPf1lwFW!<)^ty zs98&W$FFUsClesLBe)9*F?khYoM`DFEXga$wg89#Riw}QgJG*x&m z>?flIqU~aNxZkr>hN-(gyZW;)Ex41Jt5thFKTUh^?`Z^aVBQF3s!Ji1Rr!|moGZlD zXggF-=Zt}zf+ybkVAd3`^nnO^i0Lm~Hvt4LWuR%Q^mNIF<6rE6Rqn$HBR6l07kB}Y zkWh8~i_%ps($|N-ZSHIQ-4aH}!Xx(*C2T94lS7D3sZTfiqf=pK%Md3ztZy}$x|AKV z!RWf{f?c2Bz=>baWp~nRMKIZV3j<`Qz4&pign%O#S?R$ySWwvo=fB?}%}-{8Ih^-3 zK$|*}?go)3BgCKwTrxYE`S6y+-fP!?;;fqW&4!(4XWo5Z(04v-?n58Ecgu=?dfOQv zzhBL8Tin^t+5EiQK9H;C7pT6~72yeGG_d3KdWdX$1DHx1pK?BtT_-QxwyJm6+C5Ps zTX!H}ok`fIuC?Wy5jva|uD1D_kF2)HA?<8+J3Sw-=S`0Wz!XEvf_$)v?nQ(p)`lJz zKQ}yfqt*yqx!E0&cJd5>>yk%1u6J9xcE67i&1;B#|`Wzl4WS&S9m3sTk7$~s<@PFQa5_)ly?z}flv#nNdeG#6>DYya@Si^a4e@^#Dcv-z=sDdvjbKF3P#|Fc0c8EpaA5y|X=T$nn*Fax zu=I{O82_!V5_PoupVcX~%UdW->|&0P^#5)9cS>$CWx9zLYgD zM=jugz`iV8++jdK!M`-JK>iC1JJa@Z`pF*u3GhS6mgz`QIg%*OF{>hr#k)0{Y0+uP z(wVe~kHLx)VZ*2PM3RstmAxZAPWp7blPs=-90IvjmD2v|XQm$G$zy zS<)oP7q7;xp9Bog)Sis8ooD~41KqQYNOHD~11R%9YkrT< zP+f!&nM$A!G2x}&*Knu7YjB~SdEx!Lstn@X5a-|i*KbrJRMgO|J4IkXWl$>_M}YNR z0UaApW1^deOMDOWaN3%{Fz!k@2=1(L4J2$z0~}Svk^zzMZ(003-wq|+mj(}HKCw66 z#W_8F;I5JTA?T2jzMT_=5TMUoTVuiD?byL4EjE3>Vz=l@C7ORDhPqS zsT2;kfulod78`NmMKm0-a?{>D7dqsz8c#5V9B{~hn{({a3_F*JiiXV+%3=NX$n^i# zt!^8{3G*JSu|K(k$RJ#U8Bb_(-Lp5(*-U$O&GU^8+I^3ygQ898H3Bi7*JDQGly~L`+yn`d4)M6#j4$b`mk{vjswS3WQgZle%5Ohg4KXeyjk=%8)37%#@Glyk zrf-T5NX?jd01tz+o``Ba+uQ0>637LUk>7KTR9lJxs?1puew6dy!Saa4yQ^xDVmMFO zK&ASTOYN{1$aa<9vK%T^2&O}RzWKTcaD&_PC2=Kz#z@bN=(jNOe(#!Ay@vgT z0?MK+!1is7nGR;cA>w?zdasaWY(x0a#8vzdeJj;J93Z-&zT-&^9{(*xj|wWE7IYdm2cfXP#c>ZNZFw_Y zqSG37*1004smVah!cLl=HwpHyTa(EKlrQW+M`-l*LHJ4QsN1V=34TR!)w#$C*Df_b z0X?8R@gy?h4nCpFVSwF40bLBOIb#3l6jh+PjdgC!;FCAB@A^AgHcF%}i{Dn1?T+Yy zwJbdKlKwz0wHaTGDvtr4?4U_o6!qQzP@tUu4ivb$)gub22wF`Ei7KW%B;zyY~ z39>Qg+xi_b+S*bMsqo)Ls~QLLI6qT91Y5{0VbE0&NfpPjg|5{;K#Nhk*8M+t$G;*ilwi2ZHWpC9h^ zz5P4NAG=`}xDKWVvJm|=OOu{a<8I1&ZFYFx5$M$Rgz&ezrz+D`h$)mm7n%A+PpsW( z4i~zr#$sUYe5hP8uY%!T-?yMFkd{ttnffiQG*B zZwT$UfjQ`ylhwM)W@rR*5Fl8vAv2y?ZKycw=Fn%6^iklV-uaAEB~VVaM8T|(x@KUQ z%-t!HL5{fJvy$I}Us{GzY=ayzS>^YZoAV_oSYsKGwsB;2x0^-Bv)Dsc!O_x_b?~8C z7=EM#E*ZZ~$!#a=+y4M2IEJwo9jsE?f*hr#q5RrzV?=AL~wW$WUHx~y-mnr{(G4%$x*CNMbM^CkZ)Cqrhf zoT3!$NC_wm@w91o)xii#Xipb?CUo0R!yX*LQFGn-bW{{oPf|0!?=XR3B{(we3XvWmk5gre5*Zhx^&$$NT9o4UpnL6D?I43R;zsV zWb7?wmnlv{M}C=(J$^W~3|iTcBbtw9l&V<12mwmt<}4tnaWGi~tj0`Be84~B(5F41 zsjhIh5+!~`PA%0U>6hSeKcil2%WuHnKCz`>KKi;i-XNVo8U7Thx^5npU43#iO$;sE z&K0!_6ge)o%z#{hygEzIxk;$Fn9HD}l2~j7*I_S3$z`jZHhpLFnM8rN0!{IE#|)V{ z&iIVI@vP2x;IX5CodRfyC|NQUn)cM zPt+y91cA4c1BiW=n=CV$D#y@~3izu@522R0j0ZTyjP~*r+E7PqwL^J#Upa;ITBvg` z)GR^vIOkU^VFKZ2BhWBc|6^!w0{rQ%Ig5{1tJE!GwiTMTtZopS=NY-H?y#e87J)yr zn%}QU!ixv%yrM9*O}m|{TH6Jg%7)h@QnEuGF#)v*Vs2Wx{tts@Nt_k}xPkD%`FZo( z=ZokiUJmuBzrmdALwH9YadXFe9&i~oU~_dIHCi=de`-dHjjWq=E@L}*3?82rs}?QO z62)y?b*6j7@2A`=E`|w1K=y96zEtANeMdIH2u9`03}Ir{xa~j;$Gzps&TDNNJ2b}G z#Q+nC?(eCgiT=_Wm$&L8rVi1ofoh0v+%*sav~y0$r5$coe0zC0EU$opYDJ`BRV&DxR6!bVmuTELj(=sEK}W9-rakJe8l%1eczPxF93s zz;k7UEl1HL^P2^@jNT*bd{aL~-@8tV;sB;vB8~K8rXd{3ZX;cvyL*k?Q7JIwPgtO} zmzkj8uEt5m?CK;$`Z==v=zG zXmrgQ=V=;C(4`zTYASE!wPfEg{R*1l3gly>$Ki;-x!tK$E@_TANgiVXaB|T--9XGd zW<4NMm?_vxwtkU~ zzRznO$56`oEmYDb{9&&4G85g5+tA&Px3S66$@ca1`nZ5j4-La+S1{?eiK$2Sw~5L6 z=D6|eR*5y6Lt$z#d~ZpO$NyLo7jWf$fG23|`2v89J|kI-e*V%1J(EwsYvdzRFVFto zhIEt=r+&c^E^4F7w!|ik2gRW@aHZeelh(A6_VZXX3OhpmXI%NHAtQN zuu_T7U1z#avTuC;%Cl?T&h6o0rF`o?pq@}wNvIh+?)_?b6C-QF^8cXg9fLEA+O6H# zwrzB5+qUhb<8=JQM#r{o+qP}nPQL!$Q~N!&tIqj**IZTe*IKpiF~&7Sf4SVGlt#UO)oUEJ^YMZSWp5-v zjC@D`E1f1|+DEzdhP!TDt&GW&I9zORbZcu64U4E(()H?;D55i0vL%zlmUv zq8YE4XfQUFxUK~OXQYQFbHC{p-pJJOOj+SPT77US<$LD1y|J&0k$Rpui6}~U1@IzI zBF5vQLv|xZuD})0Gq0YD5B^)P%e{I(8+a%6Xswi}-r;nxfmQCE zOdeWcOjm4PZz|eoPu8He6cLbBu(TBVOB=*Jyh2m_S^fh z9&WbH=SwH0mF3ET3ecD{fAz^Hj&VowQYb-6Y$3lNE-pz2U^f$2sCq7%dLA@uV@YeR z{LXIAZ>+HzSjZ>sYkJvG;Bu@?-&2M=k*xwVh;4LI)dl)YvxmpJpFF$|0MH5R8B9DX zU51_vs{1<91tSUM9gc-tA7iuM^x0ttH<9nt1|MKx~PwkVZ zkg^cKYhgMH1fas~?&X-}0di~=eDPGVQ{Qi~Kjsr??~`0@)OnfP!^&FRPE>(8`sLRr z{jODtFL-7GLKYT)nw{f{H(A&O`q;J#C618rZIZa|*=YzL@u=&zgc^O{>Mm*DEeF6B zeDhgMIu)?w1KNP%M|Kr~s|9V*fuQzr)!}XidJ*0(Q?H>Yy*8f-;M((=mmWSuIFQ2F z!;foIL-tJ-(KLdw6lZ`SSj{$>&%yKh)39WVpBK$*X4im=_yYa zosd&U#wcs5=W(*38qQ0NqoDt>a}FBk)GC%#j$)+2P=S)s(svE|V#e`z3}8y)+iv~# zKp=TH!8;H4t+Z(bP*+VDgbSuopR+=!)`hL#&KR~1qd##;HH+JYD-yI4EPsS6wdL)X zK6D^p5{5EcMaxdhI+T_^w8)Zph8tBk5V3Lsffk1Wd+b@F>A_87I_B8Fwai^SUo_sx zGO4N4yF+_>dgWSVZT|b3eB5LH$uKoKeW|*>(2T)lx^^WO0LAmiFreB8zvjYBI}KOO zJp4^#^7EQ7$k+WbjOYsz#=DHhU|<+gmbv-mtAYd5!k%F}%hiZ_tKp&IVWA>&;!l?o zL6$XyoJ-gV+d39b3gwen;dQDj^TkZV58qeypNM%h%}c7Y?wDK`SEVZ^)HMGQ`$zFj zldPbM#5|WJKqLuOMa`kJ)?ZODkVp)>Y@iG9ex0Am^PTQ1M58D! zwE1JaAT+2Z=jkDX?;_Z{cU@W&Lu>O#I~ub;;5au-cEl=&vD}^(<@mI}P1`sZuj!=U zQaxLdW95uc5U-2`r>x-9JKx@X=r}xbo%LDX&@>z+0FBSukRKft?GcePws;uJY_Gr` zI%5l#32!%T`duMM?C68Z7dqx7?y+6e#650J&Jd%7dS#A0=j%E?hjJvzzN7F1#ZtVt zlC+b|lQ@%UGL49uM5@f?#$M(l_J2-x|GvBs0v!#-aSb+ zw{CO4Qfm%XU3(f7U;22wGZjpKvtXWg1zD3a2O2Y!70r0p4}Ey?Wc{-)rIl#Y4?T>R z08YuBzQfa@MCjB9l?#%dX6^5~eBG(k7Y2egI4uNtRaWjLywtSYmc>Y%`!{_i^)YO= zF)1~qC_D>mi)5wGw#;YTF|UV1q|DKq@^jHNGej-s!t-@$eaCYV9esRw4frDdq#cPa z9yv+HXh)0IVcfeGEsK~7rNr}pHaZ?5t2-im!lCDl0J9WhiE*8f;HI<}|tyblpIXkD*eK6bLc zJk}x>YUE_M`wFkWYpEb@TOKApwixflyJb8s&mCk6g~OSQg2-u?i?0NMAr?Xj{uK{B)wGA_lzn5!7fMa|N!M|CSh!BLT~ z7>4@f2~|~Zm@+uFxtnywmpQ-E&^U6y6i7OowK)uUevXIlTVz4JehOToO>HLBG23c- z7DIhIuJ57re9~KS>ujmqEd>b3u1AhR9o+A!?;{uqj2+p1Q0$L^oDLp%0796Zml}KB zT3_u@;?>6mimcJ-k0q-N(_OOpu@g5XX>TB`b}PAL9eD$1rVrx$FHiZ@~*yH%tCAiEz?A_gh=)I^9k zH5@SH|9O9V4Y-n~I3l#6rL)uYRH# z!Bl+@_f%Y_Kj`15N6~Y4<|zv^7rrha$@txaN^7dc8wV+f9UwvwjM?_MVBsanK8S21 zdsguNoF#&Fz+W2G1IjS`KDUkLRub@L{D`qc#5_S-jr8~{)~A<|Z`>yD{dw!$Ij90V zJ*4Fe*KB+OIbqbRUA_d!S#jJJH}B%{ap;LaSZ>hi1izHR{Ue;72V6}2hCKMdkZb;c zdV&l24hSw{q_Gkrmw2bK4p05&4Z@<=ZCG>EyK$Lmz=WYH#=->=owy1AFHJ1{y369fwXMmduX13W}v^P%R_y2sa9kPw!~zMXN^lGK#91wvlx zUzzJM!6?O7j_B2qBnShB_}Axw0lOCBuBmcQN%BWCOyZ3v2jgp_x9fzK($v);Cx*HD z`x3M?Y4}I7g&J>iOK>n$M_=BQ zS?+8lNaDZvKKcUxuhhM61%`e~X0nsxfAtx(6sKV9|Hw9(wN5hs8UHW+2e!`1<-dnR z>YPmQ{&VJC+x;}F9{_oiF_#X)g;r~iM+`k{6y=j1errJ|>ft@A) zfAt?;t>3s($v@5!Sg`Oic)v)p?hu&@qgOfwvilI^x*wS%tVi8N*zf>=QQ z*Qilel;kGy@3-tOg$cL;WLzp(^}R8%Ngm*fQEtDu)g>wCbk7zVEkHjsj(e{5nPG?? zE4S8fLQstHHPi_OjILi2m@3WEk1a?|1bbwiaF3b-DtULs>X-M?0SobjD-_vXS_#i>S@%B6LR~OcP}_%l29rGqd2auK+xFUM=K0^eX}#tYQ#61< zz|FbWo_p<%hJk^kf|!Hw=Ngu~zq5t!$~WWdyYS%Hqza=gQC5c%T4DEw0|-Od^x$-? z-si&hk9?{<1OeQ;JY2x2+HD;^e;B78$Vcg3nbCZzH3xQ1`0rQ09Rb$; zI0$J3IHhr$yW~aouoS&OqnT}giE4cBTOaK6bekmZ> z>r3{F+tQw-xsC?XX%FS7 zoC!qS{fP_UQLSIp0$0~${tJNXr)LZaSl|PIYHc23OXS z^LMc#^;--~!YYD*CN~}gKo#05iuEuL-GSkY6PT}Fw`6?{-J)&s$%s5mO^1kx5XjLn zp9EPZkwM+~ZBlJwVROX@O1rl58U@_YG~Wp+Y2rtWypk7RQ)93jK?IO==$ulJX1*0W%vy0?)K;Dz9kNf zhbD#5>O~4g8AQ?Xas{x<5H%X>&=feCWaoM$B!C60Y7WVAWBH_fn4Z_SA_BjURUVc| zU5XFV8wUG1zZYAM&@;Kxuu!hnPr;K`sGrZGjRvG)9W~-Fz5`#00tW`0S>8`_@nXRL zinm*zieIJ58e;Fk-WRS4CpbBHpq8**5ehch2{YB3u%zl)U;?0Q)i~t~z_^>;5x@3# z9lbRNN-jVL@`jPvwJh^!@Mw4OenVZo;ZnjU7i6pm?u1+!^<30paezl zQ#XJw!mM~35uA5K>XOcP*_%UhdAOI_HP$QhI_f&7jYhV$22^OCmMZi0KsEzGjql?p!3N2m#C>?GO1E^x<{TiPtbtZ%52& zj-W|lqt90!F7*T^Xkh!L&)RNd_co0JTDpa}_I1Fl23Ufi3EC-%(}ijo(-gDkThvDoJ8Cjb@w5`Kx?_AVf)7zs~Tl}ece2~9=|?* zkM{ByZ(r2l0C*slr}jk*#d39JRYvahp`mIUL4#6(h(UH%o zi5UwkM;V{i=#f-vQRN)D$aYblPn-*ljP_O#h5$V{y1vq^KpSQ{xPFTAk#jeU(SIxM zr@jIv%oj}~!H{^*VwHExtk0M}l)F(2`EgXy#RLI+#V{OPbD5aH-;#V{WvEW$m|uI` zclpuUm+tFw0*iS*>kJfXswUs9a(C>E81|K&wFo95wbZNO@DUCi=1+e*M-@7XQh5tQ zrvOAaO$gfD8DKr_=l&#DU3LzCP9A%?u(J6xI(RaiUWEPHfwkgNzEV<&BNAOKNW}Xj z-jmdfdPQIB$rN?7t{vtsA$Cif>T7Tb6inz=CcHDki?Akx@^?o~H-kD;a99&doNhuI zOo$5VoT9kFi#6I>|B4J|l8ia6k(#N8vH?t~Z&XG)_t8cpe-Q6-@${5$dkLejRJ^75t|9lnKkV!p9)QnpgIiXxk~t%e#&lErf;ez&v`md`;P8_=?2RDfv`8C+t$>n{c#Re`q_fa{bLvsj$SJlpYpTCdBM9S4gO z4NuAjT&$M_~YB*cI|hp^ON4OHj_0ea%a6ib{Fa}9^g zR%I6xzh7e`cRx>}OTL5XxL>4F(qnc6BV$a8PPcoEJ(*Ut&>}+@UAtL)iEn}*tZPopgRdl zq^*6K*rkSnTLfaUr-Cr#s9k}fS~TJ8jnqO<)|N?%Ut*nd3UY8V;4HY^S0R;mKi*Kv z;msGcb$ETz7;t{f(g1kidANeQj_a#%VD`-g?$xVg1?(EkpBtKO4C!_lE5OC8OAZPP zRVwE#fjH`viMRrL2K6+d$cD26_z&n{j*XYR$2lXJv+`3__w0{C>Zb!J?WqH6y;Y1~ zXkbOBIR_?TOx5(?u%ApWD$BPw69|}PW?E7CI!ejOQc_-?%K>7q!^Ml(Db*En%2MK_ z@!AEcQI`1z=64n%m2c7-zl|MOTZOP?;rwd}WY=?a`TNbch%hIZPuJklN}cdUT~*zT z(EO^D#2Cf%YLw(xHuLL5Z}zoFz<&iXB#!uE-Ol2d>Z;x>#()^nJm+rSW1@1Ab|pHd zQ-AvjfGYe+ase=kVFnyA`In+8F>aA=KuO-}{XO9jOlk?cps>^c`t1+G?Z5<;ch%lo zDSP7zam?;K{hF9dTV;`iJ?im8!n8J~9D>%=X!To3MvjfMJDSW=kpRXmt&K|XG=%hk zji<{!SMVdM0sgCVdjYo?>3tTOprTb$+MPyH5+d$t%>(fKMo)K(40(mFrzI$Ad&?5# z$K-Ax=;JU)N-Hq4K8Hs~8X0s-0@Cbpz7IWyMC$Zl?S0IJfzp#k8Y7BtNw@_9n{)uh z!dUWvA-sm#vVo0PW)V`4OED|D4XBIWeD2-v&~v=NJ!z)Z5&lk#PgjugskM?& zbf?Jpg9s>w_xP+EBV+l>^&sg6kQ3aF2sLaS2g#d-6H^rp%Kd)4MBWS?8_OJ640I}; zUnr==g1Z%}J;C~Cr6$S-DTRKFV())rf^#$P^5m4hs^^aKogMy)^vYNe~VOfnq1+{_OyY*cuT&vkb*)nV>Luie#_X*#T(t z^zl-vSfVfdj-BW{{^m|HdizGDTHm1ikk?hX?8CMZskLiF!_U61SXiLf+?2bJ;4!wM zv^Tr+j82+PNNXY)#i1hkVyRGx7)-MlNlNLs(9NxbM=hG>8}Bbh`XiH4U@DI-fgGPA zq!eE4GDiw0TzPzmJtb+<&KcRHP61GTm_%(6{Hvfr)_UETy4&X~M%QHX< z_9Z|F4Aad4c>cNjZZ4exkCM)S1uG_1h4vr2wQtn2N)I z%^LkK+*ze!AvABU9gVcuS)KaRZddlaVWR`e;r8%na4F%=V5kNlT>m_K=2#cxA@Gve zgEkdR>Fg=0d!jm8*Igz&w~3m=1hMRNO{vBr%kT52F4`S(Q1)e5C@MuO{hxyk^GPl6 zdeQa@tF}3=9761{6w62pCj^k#>Sc7hjR7}$R>tkR5jrF9ZM*4P`^`S`E(j?Mux*)y zu~8)gyV-#3D%o5zt3xl>S?94_>_7`3L>1SJ#B8ec`_=f=lgLvNOS{Cw-XLFx_!oW- zrKG~N-szv7uG@`=fmIRT=KczbmNY}}#@}wco=rVS!&%zwlxCTp0?_*%BYeA&$< zRoJRJJ1AP+4ERJ)c?+7bNePE^8k=p$6 zcFJ9coas!i%>zcNsGd`kV~O5XG2^zXH#V&Kur?IrCB5gWasQ_4$Va&55Wij*wviZMe z;&9gqMc`VdV1~)|2IV0eL>I6#QKlL$=#6(B7Qag;Ei*7q&M~=K_t_(D{y8XFj?UT+)ZzROsBVFTD4h<|R zMad;~chpMUKaK$=>^N;ce9X`7-p?j>m;KV+4KML}qt825)>XCe;GYvC@q^_7;=)qo z(e1Mb;TR8p^vmFGE)0=otfBy6@u=Yh=&DL?^4Z_1b^DaU1Dj`B#tF--#ogR78m_dD~&U4L@51(x|1sOI>E@ zbaQ36AObR{9WA_UQKqIb0xzOwo7ZRx96c&Md*u~x)o%q*Dl|*UZ;6y26ZD$k5v(Qx z-8w3kVaAmOE`Q84dY3-4h!;IQP#=b~S50)Frbl4@M!*`I-beizyKSQ0_|bT8oFhlS zkWJafOW0w^USvEa z!`|$Wef@2CijWP+p27`M?|Lb)Cc%v;V@|2PKRzG=94y7Z^2kDFyA7rg76B|gX@u@_ zVwslP#VW#KIKXv=1m5Dj;j%y1 zUjgDuUN}IY&V%v!IKhgZ_JMA`QaeOdK4LhC&uA!p zTP8IvszG_uN)Yr8?gou`sTLQjpyuZzb3ipU8zoLl1^WQDe6Antk}7Qn)LMLAbKrfp zx(Z944v?nNuU^7^AYT(fW_X8XYJFxergL`Kz5pM` zpI+O7l*PO!A; z5>~7|HrqJ)CRWXm{P>4NgD3Fe8sR?cHI+iMjQ%Qz*j@+}Z@`r`f$&G&Edd(!Nsc(C z6LUBzN3QcWjTuK0%OX*VFZzkhm2lBG*kWtcYb zNRjtUgbo0k%zhEE7`^y#x1}vb^ML(8?q2W8Jq%i1oAKNFJ|%YET%vFlFOagu$6e&1 zuxpUOzHE2F(}005|BZ|?iZNAw^hPXp$GuGp-kcxmHzVKN`u-~Hu>fr24v<@q8<<_c zHYPSR=s{1_lvX(|*)(7|Yp+fCW9m8s$&-4nH_aMsfcWI4 zgzz*uoG>Y`?s1~^9BQCA;I8*?aa&tv$f(h62#<9!f4uJe_}Z!X{SKkcyP9HrD7{5M z@aD?!XduuZJ^F}Obq9zcZr2)6AJ?(g?LXR04H>r9WF8DOSwA5^#b(!GF=>XpO*G1A z|9%~-n4F~hm*u}buV51w2W7C?qgJ|7OB&(bnOE5wTr_ch85OAu*L9}Ad6bK3jAqwp zac1Q`=V$qt_t9q$&$w)HN`IcaySqJYD_P@AX}l~dPs$z$m;va_#7l$YTaQCrd|B&Q zk&wSHcgu`+Kak60jmDKb8^^<+p$KejexY|@s;=ka$?&QvZSG4t&CuolD`(fhwYpI8 zyn@^TM;OeCHJVx7Oc{d7-wK%8RFqI_!o1V)^g+*HnlyhT)I53Qx_avI4|T9DJ?)H2tybup_m2?N;bxo9$<(lYo>-4k#2nC({94liM}z=x_iD zr_|a^67x^oSBnIH$MFg37AGuHnEzoN*b_pdk6Vkpjz&b`8sNIk`FZsz%iws+Fw0!7 zBd%~XeAgvyS(h46O5YwOY)|p^iRhBWfnp{1nqi`TYXzwGYCbaveRd`wCuiuYw3t#kGit2HBDef%l~@C&SD16VYNUrV4wTRLB(@;Q--HO`P|#AMa+(c$ z>2vQIAp#&g$w+cC`2>ciM}L60x>H9gpTPnBc%44FRC@3+E{JpLYA{(eJam#zT1+aLD%%Pf}_R&qcbpihdI~G)=0<>8 zkO$2@fQ^rh^oNM35#`KsiU8-wD)Hx=3}~wlw2FBgkR>!~wkV zZCt~@KWSBeV#vbEsF|w4L^Vt;+n**0(cV6%E+0+ilXH!A6TFHP-*8Rog4V>sGQ@6m zbFN-<;Sm^?fIpI&q|N)6OlIeuXH&$xX{R0y#`v@4l1wL=twOGtkLYN>V4CE|3y z#jTs#RlHVluQAv=wsXekLV2mB*YNEzt36z-Dmw~G`Nm{T-NB0Ba7L2J>@E_Nx{zL5 zVj^z%mwr=j4y;<@Iy+*EJlFY4HOZ0Ta_ltjfy{&`U~O`1e!{);Uihxv83VlirPv#N zGhf6LR+4kFwPhqeM!8wm(Jg8sYB-w%!N7neT$JKv;SkI4mcsuyD%%y;l;l=nxZ+Sd z{uxFbQ;;hr(FVIv6VGFx8VMnnY~eYF`&i(UaR-o#S3+qebgY4T)hSo%!#(829*Fy9k$>87> zC*Fy8H(k)l73`1CT@vL|vtT-m_l%QQldl-Cx#fRWf;CUL%C}=vTS$ejM4>5+`&7`6 zQc%yE|Gj!IiorG7RA>6L&g6_DX`gVBf{Lyr4NR6Q#-&4Q?}+kep%oB?8|+t#+>?Z> zbrZy|aTi2H`R0781mvcqOpZ&PhbmXAg&Slt1ZfSF)SOI1@P{+~`I*^S?oE-4 z|Bh3X{K1tq_4KImk&wvNa~Qn!*OBvhXjOi4K+oizC!0(YSf4 zl?xaLOs;=#acjo3&)(xIioVd1iMqdp3nT7quN6dUmKk6Xln)?KpJKVeg+Z89Cu;1zlM4xcoY>k7ru=bQmrp}VB5AM=3jz?SpmO<;8n5iuf|emoT>9s z7Y8odMV^V+IR!YC_swcGRka=|Iqu#zcCkyVxN!`%#~##TlHy4MHA9}1V(M>I0v2uO zKMy-V=x)aJZH+m>P0zMtp(0?I_K?nI?uj}7(haHgo9xLw2@!U&Muzuw%M_g5VuTto z1AQ@Mn5w-6vJrD{ep{W8H_=Oy%=Ke#i8oRV zse~bT`PMO7=N=R#j!Q=2%+dl^AE+>_;+ORg%h`x$%^mf&J^ty>Q(cZ4ttm*v>7r2% z?PGaQrvR`jxl1ej!}btcw&i!_Qj`<1v&S$_wYARJjs>>+lR(q_&g};)k}~h;JeC$L z-~W(^60R_1{VP4K-rBgiGrkqu_R+c0jXXf(=ldmROj2<*J1ak=Eu5p>9P4y%FV+ zoURJ^rIb4?PcJFGP(m$$`x{~y1F)vGTa-@^{$8ZOj1d4WF< zc_(IU?3+vQHw8>0Y!6AeQ~X#dhjI`V47SGlA|D$JG#2b&A{Mp>pqr=QJhIqT(gI-p z=Tkn~kBxdv0+v`=fqoD2_44k$K%)_O{80jt&wI!R<05vu+(Su~Gr}+wXdW44tVw(6{@aW@%pXyNsZ8$(A##F@SEf)&Mxg>p-$F^(ByP`DaamVc^CLiY$# zeWK9hzH-6}6#=*&FeD9j1YkZ?kv6&94Fk6B0*dpcb9)i{6_R8Yfk15KzBR$3Y?XCr z&%hG`JBc;PM2(E543n!&Q;s`<6CV|~IKVQE3hLsS`#a8O7ZjXcVwcHLmV<64jl|7b zGvCv016Qy0X{alq?-Rn9jW406$&!uiqGW3y6+qVE`Y|BI8_*tD4Y1>YX%j!-&lR9` zQ41}+ZVA0M`vf(dxgkqnwhN5Qj<(zPCXI=)(3;zqfC1!YxRilHh|?v7JR zVM?7zMk2d3OviywfTR1st>-vRZW82?5JSDbw1*vMK*PEu}44qV#Tmu$0Fb>y_!pSNT zFnrOOl{queu|6)qFUY|0h_-6mK50WQreA081Z$~UVTUG~n{|K;F|~#6?oj{2aP-UR zGFMQk1QJie380NUM3HWnrW40TpcEJ+E^*`_V^^Ox+SD)!Cn0Mg^|#m6j#x7vsu!)T zyYcTP^ujRJV6VjcaaDg9+GIX)CE4LP5OKckLeZF?$91R8A8Ma`eM z0wnDsMFCyu_JkkGO2vtU}dColCgA2oPP62H21ETz^PAD2>@^JjUz zi_9!2#oARW0nNF$L-`2APd>p{pdsfgBD7h-!NBqU?cBeT1SWQGr1z}p(KPn!+XB1P z%Zd}HzW^-McP8AH(Q3&eigQklOy8ERmzVh@mp)!EWuz`ODJBdJMC$1l#k(v!1K+vR zFUUyfP!oZC9XVh!FUzjSFmTo`)R$u0=|~E9{$zP@%&@=f;8Aznd5^4IH7&}C*;;>$ zJRzum8OrBjfqKP8N9Hjr?wkT{30#}F6Ibp7U#Avujw)^E zH)cp3So!-8st0|3iA4m`N3$=SsEL;W3;^;qOm-3HGuK7M6RPsUVO_|zna2zHA1Vpp z?NRb8;OFyE>e+pPXJoK zi1>_}GKW+~ek&jbZ)2(PM)v5CKxEe)|5%)i!QP|z1$3h<@c08`b{X+2{AKu}g5skV zCdQ(hgg$QJzfa1^nZ8FC#v1~eq`Vys>ISCppQavu$#E9_9D&Obb0 zO%IAwctEQr1^!9S^FPlcfI4g^Fa->lTW?%ENu+Gidlv2K#+Ergu)MnbX1HEcP|-VdEiNq_L_Y4fZC$`u2)z z7d3Nsr|ET?0D8IH(R`1@coCuWCRw@`e9__&Em7SKu!Sc2M0GF8qFl{&Tmrb$`NWQ# zOiV%FSNC5>fE@~$eFhBcdREF&ppL{?oFQ7NnT}{zEo%oQ`1@?&yC0r7?-2_a#_HDI}oLE)441-)6Fm5t+O-#3OXin3FgbU+R*~8&RACzQcrBb~wNzm;=8RcZ_X9q>1e`#<5HKL7 z^vr|zD95Tg|6TS9EpX0$bd1yyDGFJiu*KgY1=lb`IwBsf$Ufm_QSZ~^W^~j1eVB=W z(LCp@BYs(wT`_&hqr|5$RFs~89*@6d_akx<(nK?eF(CE>@cM-K1Ny&sE&nfdC55M$ z1}g`@xCs>I|C710PUo!pUl%PKz4I@^|9EcAWu0mNGyWfg?tgeM;JjBCDebb((*N^B z3ap$nF(6=ifvviuzM z6oPD30T&m;D!(J9vT~9kcD#QEfZ`Q%|7Cik?ERjAByb=h&?b`-*T2w}7Im}bfPC5X zmUS{e&=A=WcafQu9?Fs=tx=3pdR+-|3xs(q|Qe-EDIndP)!cy!L&cAGO=LT`Mm83{I8&6%~*u}b2?!AkB`tdt^ zW%!V86Y~GC=o=(c$?3^!G0g)#VCr3J1N!F>$E7N=!VPu}=bj2_w=IC*qgZw@rH=hS zcH56>n>W-<7!<3q=)vu@Tsf43n~ZyUGs;e6nZTp-EDdT4HObR*>Me^FkkC$rh}G2~ zdIpe)oSG%jd}0mgE)2?9zMPi`^)d;3+NO2&_w`nw1JPj5WcgOqfH86d5R`})03|su z#b=X0oi8$d#a}*m!ha-Bo$vBV`Sl>J@Y>Tv=-$2!1>m#3x3>1Q8!QSKWg^Wkl1FUl zg9SHEU<#?`QJ9ob8ET_fAdipjmjYJPW4|;ux{Y3YXiKHAIBBF&H(V4YSQb+U!OJT% zsyO!VszHH}n!D8Z4OnfJp0}2G9K! zqUGk_nwOB^uoCQ+m8R*y16XlJ29)D86aUgLzfGWbAv~Coem6D%!+0EQd1V$i?V`*zAhRe>l;c>eXJ`$7_mZ%8j+9ofria!JMgkZ5G&m0x3U4<{VjEF! zPcta_-sHAo`f{i`0PwDI+m&rA^BHv?XpRE&gLNtK3-jUioFs@9N3#BAS+

D{tg0}>-@pADu-C(J>i z+n(YewdB%8Dvdz@C|W=tf(V7OM~x&_kyceaG_EWXzKBY#X_{MV&(J$n3|8#Zx2U@y zYhO81AfEBCt`HL6r8j2@MRS3kPewAs79lfz4>BAX;EOQ8FDpk-vBPY?hB2YoPTsQo zCCWsj3p3b=Sth9V3|-btTfYo6;TJw-fmSW;Telc-Fc@Q$r(h){n6ZwzDODe+r_}cr z)_5F6`U_2JJzRWZd=Q2Vn9rD|JsP;`;Y5F4ujabk#w6xnLZ)VFP3u zbW68R0Qi)8enz|czqa0YWvo}mM^`w3Kq3Xpx?t;ODaOx9GmO7fLO;SI1N^PpkvV!S z8mPi3P{Dy+vttpztI9qxg=r!}E?7|i>ZTcSK#4CqR+5QcCZ2u`N(xQ=T}Cfv!Z%FD z3%*o8u2;wIq5UnwnW}=`^6!m}{TI@SoS+XQ2v8NIF-cW~O8FNH#46fbEr`}|>KMH@ zk99#(kPcQCyMxwA3u)pse?C~5+bDqwFe%4I9)cf)-!XP@7tX7OS@N}Y=osmKu6(79cGd^+Cr^h z`o{dP>>%0GmGQARCR9sMs4>FCw$#C&mFV~JnI9}}_M2hU)MnL4l^w>f46xd(vM|HAFq>OT25UYRr=Z*dqeM8;WWjAQyv5a!2+{OlXT+D1tmf+OuCq_6_O7qT?8S=4(`RNvBAgzwVL z64wPiA7_qUB*I>>O;W1@tA{Z^ft1;r#l2Ay>N)a?HHKD~nqjDcx!i6< zvrzN|wnLQ%C>DS7~4)pOE>KEhK@ z=TnnCAU`r=45%n+kPIIOv6EO~7bS8mtQW-Y4NV5dYjG$n#W}NM6(ym6m*>T9U}yw* z1j|*Yyg0Je5R71Vv#$I}K?yg(91H%j%7V8Q^xHPsWa^cAu+G9YyZr~oDtscY@X-h7 zwJcccp%uxQ)iBsM0Vb7Suufaq9y83Z;e|@b_Z&{L5l$LD_WRA8Sx)E86inujR1(| z&eT=*n{YNZBMH)AA?ir@OL~81{WP+fx98eU34VMu`wwd6cZh)9qHrj5jSBYkgA#o?g&NVE>ZHv#1Ye@V~?=yvbOQ>S3i;tmY0a9a)b#sM!QlA_$E+z>V@p=sJf;jFF_T3>z9 zPcD10cVZrRLTBT9hYus>mr}OK?@D--iFH?J5SM_N5}&4%H)sM;tE8{^WTq1K^LUe+G`f;Q$>e|y>wt# zB(Kp>c~o8x!q5XSI#W>;Vx~;&glCB`J?_!ULaXu zGeIYJ=!kv>iNd8^agJx_o8My@FnMYU$>gIw$C3{o+Z0pnz{qt}n0`{QMubjTfdI2HutWvrPAfhx~P?Ap(XmVklo71cql->0F8*~#U3>WOILYDEa16uxroccwqgpaTr_ z97IjgvuUI&1(b~aq_Kva2Zd?mF34Y#TmH&-i<=XyU2k96`qlFUGnD5k=W31xzqObF zs5O>E32euxMKs!Eta*RqW&M$hcEiox5m!!L%Fj^xjH9mzr>foR^nU;P*k7_*M+u!d z>l`pliq-CWhX(JQV(X)C>3rw$w)$?yDeVd=iOP7W5Mo!l;TcuDye#{h(9w>?L!7ea zCt)Ro#kJpJ?=VRbW~UTJSfOpsJMj1bkkZbs0BQz}Tv9c_lT#LLyvD?je*Ed`AZbb| zkn$={3sK1HWO`U2i*GMicabYxw^m8Pc6@e#lXLj$Cp97v=`ih4p| z^}E~K&#D3fO-;n3iA**lleJfUP$XnKFl@|ym~cZ_-RNMy8B!V#LTaWOj-;weDi<_~ zI`o^2HZiRlepn?y6)g@tk`6QyI*StJ&f#C^^m z)vKb*RZ~xF5uG1Y8kb$ADeE~~1G>+mAukn5e1a|{$F%Dz7JsW$^tL1EIgXGinUuq{ zx=Nj71V)SLqu7*F8F8_}VCph!pPjm&4|S-W{UIqS(z;;?nPC66-3xLJ+J`vl8!gJ< zuR3+=ja*K?h~2J4X=@oN%q6UqeFQQBb41eQ@7p@`;0E=ZMPgTEX#cA<0I(-F<~v6B zArGx%0wXGC@Q)uYIJs|Qok|r4z+)^Z9XFZw)7sCHuTJp zCrGvj0iKYw?D4X;Q+z5TUSSAthq;lTF2f^6S2u#3i;xnnf&mF!1|$^S8ZQ%HSlPAF zD?3@rhWe082Z`=;9x10wP*;{0z0~*@mZ?R_{w=v~YhN;{S<*1NR!NcQ9##=E4`lP* zlfF7xWSjSPHeV;eM|4i!IE8j`?G-g}y$r(SH7baD5~Uufo1@S@0zD==4~&VGjK$5i zN4~uvR+*?LxFSJ-G&3)Xl0UGSZ>!-RuIt&+LQgjmr*Za z#0`7gafGpWb0fZ;Mri1q16U#YQyoHYN3V?Ldn4`LG*DCkJ$9QsQP^`w)Uh^4ph@HpQZe11Ftra&{oNEQ#%gSte%~NM?lr&(rL)LJ7He5 zUtF7!G`$0}ptU}C|Mzxd)1BDI9)m0)(yzuALL=90fL8v#T_lw`3 zC_vMSPV_jxT7{J*74Nd`&s!;U{x%)ZEo2vxz8CeE8>`E|>-qv|qc!{?;fH(E1W{#l z6Bkab!R3NnI<^}W3jknMj$oqGdF96tql0G)1S6+kh=?wcGoaFG&ZvZT%ZDbm_W3w# z?3r`(`Wa>-(rLmqQ56ewQZ(;R1KNvS=v%j=o`T_?=qq23k3+t7KN1|)oDV|~7xP}% z*%IMzM*tP_)PnUa&Cs@SWTW*28G8mNph z;T=W=PBYnab}Tci59#LnkDcs`!0K>Z{b0I$MRg^w;cSk`28LC$3{O)k;G?P2wu}OL zipTAz>YwV!=87J4D0QTSX@OB<&2p|&%XK#kO6E3hS^%f96PAyy$Zz9JALLG+r1Bp_ zf0t|aRYiPJPPU)6a=(Kd9KjP^rVEQ+zzJ*!8-=pN<>XzTL~r9BBhjxT-BX0_{+jn} z+gh#ua((HYfn((u8g3JK3qk`mH+FY$1Bf~K;n|z;EOHe zAsc%U#UhLG6-<;dOwEHyh4hQuoq6l#Bu|qjp8;u&31P`{mUg9B!tyutd=ovN+dXgZ zY5JVez9e3`U}UGRLs^#M&>6iGYS`9)v7*~eXYh0`-CP9OsxHn4{6s%7X^8W7IkX6wthL62kUX9S9 zRkejc33|>U)!f34=V1l4+iWA1Rw&3GY?zO#S{;7hKx(n9Wa4g%)?uN>brqlOUI_g=03^r4N6PE ztnUE@O$qNF901-*f=#Rq8}PHTl|_YAs=2gY=KCLRnu5PpH2UDW=952F{R3df6wQRy zW`8X#2-!j1T_UwSt8~0I^Na#uVz{ky3}^z;7HM3fdlVKl9(-tWE+snghX#6W5|SnI zq{R?9qg90aib|5@;YF8dp1J%SJ!IAQxZfVGWMTMGs+vJ>Wz_uvtjua46-j&?ZR`z+ zK4Bzx=O>MCd`?K?-r|SYyCu5WfuK=qLhnUB1i5dJ8`)Rqvt>KS8lWA?X)HGO7exz_~fjbojVHJ zOo!Z#>?s=^!mKgXBw(0p4-I9-x*yl5y^Rk4T;q-C5}-lIw{oS}3J`ZK<9Hxo6un_PgDGN6?PJi? ziXqNMZO||Sl?f8?@K`>k?&Me^Y<3q;2(XMP$B%;FfLy;a8nfUbrB%{YIu$z?RJHWZ z>za``Gp6a7AZgSkp%WfDHa|SCjJD2AuGpuO8>V#ajk;6R?$H5*;fNsfqwdMERCj9d z9C7>Q-`8w#up)kP_%J~?E#57jppia8`<+!>3bsuo2Y<@9Sr71QC;md&Vbb-(G%}<` z@u|5A&NmjJRHupXB#8%S#Gc0Y{`GE6?uuRCO^zt1?O9Y7!cN`U>C*{o9-yHCz-_b5h~ zsLQ8OcTd948C`MyTF<^u`4 zydcYwQhzz-eu=3qxxD{*q`(r(vdB?b1F;>pOMDDe>RlxOUzU_L@CZyTc8Q4OnBCY7 zBPZielZSdhz>x~4#0p}d%3Ope?YJu0cViH@OX<$QNU3uN2A5vuL=gUb;W5?lOpxEW zS=pKVRgP^TzeVHA#?>=r*ZIgt%4d=sjNc*|u7o@`=E<5dr7{ZyaJN(<)N_n2@a}_* z=8XERe*H`V?BVL@QD@e$Ix3R<76V1{1cg|_HKyTWcr1f3p)>QMGmnf>GBvvcXF&3+ zCN!owU;061~0t0f? zE#6;DUy~`iVQYSmi z;Ye*9P4lPL+H0lp4Q{Mq{sVVDw#7Kqv^q{a0ZnqBX3X!?B4QeFnxYGdaFu5W#{{}x zRMhICbO`8$2~I*G88ro8-LSd2EUc|KAtSw+5~Cp)53vM0;uEPbUsln|-<0S{Sd&^= zo4+Y10QMq>wq1&xtgU{E8f2%PoUa@a2@>kTdyPc9Vt_HKsaLZhJZsY!&XKaYLitRic6UrCEI-QU$j;q|E-H;RM;EKVpQ0 z^ulqp7=)+T$Iq->ji-TwA=FesL=%iI^o)Ts0mK=X2n4w7RNfa3<=}VkANzjt{dt}; zS=oe$G|^=!d)V*)q@2<3cj(d#!FIW;i^JU>uI_A;Y)5HMWi~i8%5W_=Safp_G$e^A zWpC|RH1~|P(nBnttTz}LkD6k#t_J1e1oBre##ObKi7HxX^pW8rgdTU{{_vT4UjL~g z0Z_N8`qfjw-6v&hqR86OmK^y*fsZ6Rwn@o2sEKKE29`p}G26fv)P=$D#3b-0mEdZI ze6p6C%_P+ZnO$I`uux4O>1bja*BN1dWhBUu^<)~i&+Gnyy5d&{m!7!NED%>?P%gfz z^ZKiGno^wJd@XxXG7eN3Vx43Wq7qZ-3D5vpIG%%~fi#YKku-R0k_t#7RFKLxc_iGj z`bpDGkZVl?M74D&baKRU=HD$+R~mr_3+Lr9SH zq)N?J$@tjYjb<~%QhO9Rm(V1ov2JTgPSAC$Xs8PY!qvq-(maI7L zZ$55*2D#mob0}D*K@!4LafZRBHvm0<$is$()=euz1YVAaGa)33)F^Ba#ov6`m1Qj@ z2)meoyFhHz$Sz&sd>QJ6@-+Z+gZ-nnG09akjvSsHcD5=p)L({W=J*Bdq zT&;I8 zW<>=(vTpRA=X|x>DP#JMN{ZHR@6WK6jx_9)x{1h)iuk?S_m7FXK(Z=qa$%^(^rDZ1 z3?uwz_Pxd)oQX0Uaxw`jY>0<^Cb@c5I^hVPRh_E1;t(sTtw&_fEI!KW-`3jIF0f)q z*3kD7MBn`qj!o-@pRs#` z65!g)m+?KgtZCXR<|pbQJ^{#WP9wuiS{!eKiJ{F5Kcofg1CrK}?fr-KH#Wz3raWsV?WZ5a?BkYrF zTFOmYl0F<<=46k?dGtIPv-xJ~e(4M-*lp)ld=-L!ALz|xXo)@-12^#gbh)&>Za+`0c_jU+m9hPK zqU-l~DZhT@(k0uV_d-_-g6E51Va=pJi+d{XcLdti&?3rbH>W53ZGcDcMoY)t3YI0c z{?OFKZcRedA8m_7AMKMEPwOw-qh52yw>dSkaY!{LpQ(guH?K{F8T&bRA;p@*=o@fbw&Xr~FFxk&yOi$-(l zM*%qWnmaWg5-69@_YoTB*`bs1@wdzu5++MxumN4%VC+SE5mMUy18@;yi8|h~96!Dt zo`rb@ygFA%h`9KnauaST>`8|0C;5GP$Yuj?iX9g&(k{YNKX#>Z%4^00&T}#lE(;&# z<@srSeFDTcR%wxz0?S1{2U$=|6K_S{_yBZzJY}RZC%r?PDFf`Lm{19Sxf^>&u%rV^ zmvZflvL{__TlBui8u4k;|GJ16E{b|9F5!NeW-HMN`4M@Z^)|JX_L8+x*IFfcb<ki17l5PVr_ZJT7mNqiYj*40Y!G$l0vL%^Vcq zlfTUsDpq?Rmxh16n6wUtnY7#6f`yx=x;xQt(m)+@Llx=FhC*r4XF=QV&N2E1)5 ze%q_uvFsa=SM#@TLT|zDd-GBIJr5w9=c^&T(4B?z<&A;YqrGtJ$a7nBMYvqLoqP0W zM_b|pZ)n-ALHm+sC!RzETawxDVtAk2d3sviol~Q5bdkP^dYklGGW3-$gC*z&@{VQ* ziY-}m;6PYtW0^fIdy8t4X|h+6jzzsUj58#YRf^d(Uq%*x5_~h1_5}QL#RS-1&%u}m zvzL~)DWQ#2_pHZ;wyJc-9v8WU3lH=3*eiOTi=kluCKRG=Y*coL zfJM2(ORo^F8EL26QReOLX*sTz0$T-k3Hml6Y)HcS$=|Rcg#bCQJ4VJbYL@kLBAMaK z?eO#9z@f9GVJ*GwXmtL-N<<}qiQD8;apUcl#F*cD!5x39f9C}%z zwJKiK+3Ahu>Yk|f6z2faZw?!wq_pjKZ`WW;ytQk|^9ZEgg6kRK?2((E%@|2#kPbA& zDVE8RD1&@aUG175mwi zj!rSE&Q_LO_Z@C6B}YL>ZeNY~uNxe``?1B1d-4FXe9ky2w{?!r)6wkT z;EmnV63)`}qCsRO*dCAKlWgfGM)+{vY|IT~kzRjYdB5AzPdY4woJw|=8BaK2f6ksE z#Q`hxiR>iA5Vnz2v=Iu>l9q#U93c||s092N2E4Va!7E=PFU$3g-Bsa|nly_{f2r>`9Z(?DUy1 zqgpxr5hKyeQ@RZAD``sTLx(D-rZLgKHUR{yapi{kPRE z#NhB{zSRqST3H!X-i9AZO~s5yO1F|;!)@sIL`pxQ_E#q}XLN*?kg#g`4*`MN<(V!wbev;zm-4rx(W}5BPSz3;{&?t?vG|2V>l)9ns^!xO z!)t8fEa7u+_mBL5U3XOz8FOtM&CQM=UWJ#jy;9ciyKpl)UEV;z2$l)fT5=q|XQ3;f zzQ|-0_<}z``t(nYyfUOh&2639z!6cMdgUo>$qrA4T_?HT?oWZg=Us2LYKp^a9JG-) zHV@GZbOOBSFec)t&{Hs;d)Y<>LtVliM-{G=Xfxvaw{Kdqwpek z#__|gDYzoyCKw_oRTYvh`F+djAi=Z3M_9S5)a>^93!uiI-%45Kuo65e3XNGbH_Ds= zLhd9{CG%4&12{5q`ENSA0&!@Ae%ThX*)U_B#+b6r%1aB%J_48Ue{$0{m3zLyWSxy6 zSqUIZOeM)tGn<`*@(CJ6-=7*#Dc^18bA=74;6TnE=PUsm(<1HzeA zF*af@8I~-MpEsP!3*Gcgr)qwCNV+1L*`=8wS5piL!FmkQyFhZz0{7J~MMjPZc&omw{^7?6Q_K!xqW9Hm$98ZhUTxxE@KeG|JaB~{roN?7>QO(Js)rpMoy z^@rn5eFu^S5sE~}0C=fpfgV$e)8F;>02YAIUNDqtj=aG34!C^2X*hOc4Fn4@_t*h> zu8pqZFC5UpAJ8trmyKm+V1C=-0G6Mn^26Ux3VvAaSvU}i9V>G4^5N`-kN#K;qz~Mu zWl)`*J|_2`V^Qt5Ey$><#H>Br3Os(F%@YU|)*%A+A#%!b)08i)KwWh!y{Ks$o|9=} zAV-I%D#U2+HsAuEO0I?+c7;i9~AC19qXEp{@*4 ziRHvV0U-sH$;#gnj`z8v_sVhk+BscrYc`tsR z?^aT5o>tKw+qiZ(pc6Xn)&(2==J8X&A`ZB+X zq6n)s%wZaw+YQe9joS@s0m6Hxeo+O7LGw#j#uS-;seCU<+SS4MaP=Kdp~V>5Era_B z2-o(Y{vJQ25lE5`rdCfE!r!NI2#_j=`$OFwG&4EUX6s2@=9w6e36>leQPU$|50Q0? zs4qAGBs0+|w;WYQH1LzgNx%6O@7oy3ZUfckeDLx;{IBjyG9mOw8z9!qq5MEp?|cTO zv_oOjcQ~Htd>2okXZCD0V5xM@llfw}h$`aa4}u08Hj!(Hsi4oGW2hKh z=hK<+4EUQi5k$tbR`UDcuAyx|nEqau!tB}8F~;P1{2hp*%P&FDrNQCuya zM?NvuDsJH+V`Y@p2>{qpWcamV5@&Nuki1q)gsbWv9^S%o=CQ6RfU$}>XSWHO7{kXD zF=k)0&F?8Mm|pLl(zQu`DR1?qOpaz@geJN8sga#G26HzNfE%laaD7tb?*6f3&qlTc zcaLqKvmwqc10tsanx6Pjx^l_vwiUE6tWXgT&-?<5ev7m64A_A2PI5D3mpE=QjoHLbj;J2_xHDrs;!cf?Qs4feW-IcS9nb4^Tu;v; zxrmcpr&wj=8z4nk8j^3-W>E@=T~QKVxL2Np@gu=nNk{ZN7MiR=&)xeb!VA7fQrGqE21VO}>dRk< zXBwxBDyKXc7`rA-DTn5Vfhd`U>5qaoljrS>?F{Z;mw?5+jY(?725YaC;K_$vyh9M$ z90I?+_`Ak!A%v(24%G$2=1&h>PW99LsS&@|1ldfkW7s0&OB6V{*v`oNA*fuwQK~Bb zAO&7rdGF@l?4Zr~LVntSWb-^ruw;%LZvzEMR!{j_lPy5LMEWn8a`dAqAgxPibUBBP zdbmnY3IL*Zd|^{YW94(wKZ`Gb;P!#7k3kbQ5&ndDIKbw7@Yntrsq)m$9a@EMMwg#x zWU^Xty6cFAM0a1485h{D)jVn=q<*-qbC#OfdD}1vW>iG4BfXI5BcV6wRPL7+U;;dH zC67tV?L+y{_^Z0~=FraHL*1Pjl6dA_QjS9~Qh*s1yJEB|dk4m}uK$>p6v1`xvIZBz zuyDs6Ifsy@u~m)p9<)gc<}58tAGXH?p@(o`Lg>?8-9y=Rm>genAXcug&qY{Fzb#eM zm6g-^%@0>!S`9~@IbbbjFJG;H!hjR9x~Vs!-PqQEE3XBtYLyg;C&KAt2)>N&E9(Gl&<2zQ%XmueySoLQLF5II^Gm#_oG01(|h zd8*1#AKB~6$ORq$lo1Z%5liXgEO%f}iG`LCl~_wJAN)J^#2Pr>M^l776TDI|OKamQ z)$-Vd9$V!u8)G8{UIuqc4*dqEL5jI*3sF-T|(JvrNEnWlMfpNq)3X zfrQl&bG;+{9;Etr!fh^b?^oS-7vbSIUo*aMYq3A@=Cy+eS(b{b3iW%PwEd5cTWy-aWEufTaQ7&rB%s{*Y7-iu@6*@N;tp&fAXW(IaO7y+(aK zYHvPLo)(FEJVM)Ya={2I7Yi`MzpyPh;d)@@d*K1q1x0ZGBC;IAesZYH{Qevo-V`M4fJK8sYxWo!p9B#$-p@RX|v6+(ZvA zH$OkK8^vWbPm%HvMGGLUMg6<67mbGsKt9M5wRJfXis7$-7o;KK2vOWPRnY?`!Sj1Q z5jci_P9;M-Z%K*5{G@|pHiP5We2EQJ= zs40f)Pgo;Ovw3t~qXBBhz&PgP14ekL)~N>6!qY|b-(jnBb*g}^-CiEmb)wjjh1Py^ z6*9^yvm3>2->5KRcaK5XKl_G?@eF&<6}RcheDb$3Pz9*CuA%RS77#X7X8UiNC|M;>fTA^)Thzy+jTub=wgcJQM-^0^Hc4kzawGt(0sQQ@`BYA(1>Hw z^qE&ZZFBs_fwBakP+yJLL}ZWbmOkNbF#%&81!tFebSmYh+6Gr8r71v%K}JU-3l%7k zb0cuhQJuLoJ!eq)J)x!&$x-gSs9kIU`ZyU{VRBH~nrNBc&x&S;f3! zhS$&L)#U*)*KOUy_U$m1thyF;GNNUzT3G5Uqc_?&q~1N)an%ijZ}}++qsV2CQC-U; zcesU-ccePBYq>>)Sl}k%<|<J9dm3{geaKKK(z8Km-tpQZxZ z`1dQAmtPc(aFG{ClSiwZ!fbQB%Tj#%$GZ>OhQHhT`x)i5+7olJ{%4!C`j&v`2( zbCPY&XdzKs5Aq8psR?OSTW7T0MxO7EMB6iy7}l7T=rp+brF>dP^MANIe9ti+y{?oC zrv?Gg3hW)Gfx&?3IdzTog*f|)onc|$G4u7v{El|$S7H<(V^N%CD%yL0Arbl06Ur%j zS;3_NA3Fj$G{3~ftkWcyH1yJI*xYBq4BehXr5rPI-VaGb7V9|O1*98-`Y>S49;njG zY!{WBQrO;|f>g%*^&^h&ZNt{5KhNjyq4XGlUw9U7AV6POL93+voSl=!n&ieDjs@Mb zOZAYc(p%D|x5OkC4Re-K`^Tyl{kTnXv}>!pQdGuDJt-PgaY4pLqnKH%Pzi)10htxs zlqDg!9Pc#-I{{Qz#g!mUXp{D*(!d+++DJLp%W=^z*`C`~AxFQaFbZdWJ-OzN5@P^h z5ZAJyx4xe}{zm~qwGnA4v5|0%DC`37I`|ZQCJM|B7%hQXj5SfVX(j$IJg?gzzo0aQ zvaFw!+=Z>lgc|{xEX}&-$>^udxLb?*wbHObs$cuX{N11vUR|+jaAKqM^WdS=gi~N-*|0RJIO)k`?3q301j*y?rGJnmckN3U}-b!|CBThO&VO z@iI&*Nc^pxB;i~wz8REktp-uy3pg?FQ3~r=7`3fF`u@25=Fo*4C5<0i;mpLx2GMO3l zUxj;L$zPYMn;vBl%$aTX+h^V?PUG0WFhXiOzdFHF47yx&skSg##cI8->RQO7h~44Z z!{;#8JD!C%N6BZ4CWTQ(srx4ZK15ftnlS=$MkX!VY;XjQH@z@m&v8%U$Fur@{Z?k> z#+6eF$BhhW*(zgYI#M*yN&r3qN9>s2hV2StF&eUg&x)y* zOP{;?vSZj?Y_q!@usepiUeMFq^Q5!-#)bU?u@IKD+1DtN$b|DzBLkR*E=B)ype+J# zQ!IsY3wfem&tAV1BMLt!y{_Vv_}%!V1w$hl1)SpzK46x>${_m9tG^f zs_vCDVb#27@gM6HBL;cS;q?hzz34beqByOd(y_fKYy#XFSf}^Zu`zWA=XAZ;y7PmR z5Jz4CBr8X{Sa(xp2K3i9UJ?p*I{f~lR?eYPhz`S?)N~^uT>C|UUz+$jT|CRP{zPE3 z1=CgeLF(s)pSZG9^lJ#BQf6YFIhXjuPU0fvs_W7w>0Q`vL@|hO>gV%BWH$JhA$nKY z>~zTkYjAX=7HDw`axcBC1nMo?uZO$jDkbsPAJHpVCV4`Pd^<5}6zP$CdhryJoL!KK zn)z{j`OV}xm!R|j?NipWd+=^(2T$TDtkoV7nC|;D11T{n&t^H84=1?MgdzI*5>*<8 zT(a`iG}6>lu3IXZ%rQCX6#t{lo^7f~?YaZfaI46uqC^BoGn?8+9#ZaweFMGJp>pLO|s zjR614-{eOb^x*j&+uZ%0!B#qXMVqeLMk;Nv;Ze{9w#G1K9lyXzseGkDxiiTFo{2ba zh;jbROtrWVEGo(`^LE2JOtv3oUHjc&iu_aRnqH)4vN+-?*oal-d769}+Lx*%{1Qh0*?MnP z8W`;r^kFeuI|QIsGd5=&W#VnWfCoywIL1TvYngB{Nx@2TsIe%jks>Of&s$wl<27_3 zs;uTn>N4%u`c_#NWkQ6>U=%?X7ds2O;E)!H;U!l9bf+e0Q`%&SJiUKn*Hx;ks=OQV zYzL!o44)Y>NtVi#aOC??x0lGs6owdf)8k6&uzzH(@=@b#_uKsF#cdXf4kpL8h8lo_ z9$|ld55Zo=xV}lkRni!;aW|wzt%qb7RQYc1GkVWMQ&wI|!At<;b#zyyW0J)m&bgWX zE&W{_picqWUxEN@9X8e8*r+mpC2SG#%C1G4Z&ia}WT2|f=Pce3LBWCW9q`Sv6u$W< z1L#mm!*|VM8GuNm$Y;>HN8hv_2}c3`YS;H(S?=XpmpGjg=*KFQY5_YZeqI@TG=%SW z-unlC7neZgNWAD2(&PR@-;L{1LAkmy0oI~p0C}lC8_B0LO<+@g8q;m)owh>Z_047~ zTNat{*u}_YbF^?h1=NJsL{1y5j3H39g9bizCy1XxNqpZUo<9BPh;8PMc=-4A*bUJJ z=E@4R>YllsFGU_`1pe5*6;%Se`u7Gc1W7rOH+Dg2w;5EQB&YN_Cu%f8N;5y-w4;fy z0hORjmTi2!DdH{4_adaYOZP(Iu4?isn~BJqt$54yb8D*woF0Nq=KWkOYPar2fnCwR zHK?(5$uP&huW5Zg3IeibI0%Z76D>y%v+`jR!!}ku?9^!y<4E>6PNZ<6tdBIrH2hly z7|a0U=Fylyhu1_xq492O&@D-8lF}h!O5{goZ zM9uN`i)GhPc-Y5v=WBZ9w1JKYV;uOo=KdRcL2!9{ifHsyK*l{Ql>% zjo0V>?dIW`pq}67n)m%c9Ka9oyh-GTD!Yr0Hq*%4pfp{luxT|;w`zSw^m*pGzNz=) z&D>z)tRKc*v5X@#5FmIR9 z%m3JpB-s7&<4%_~22*pfQ}y{6Yir`l&*%4f^V*gyi@q0n-F!!f1&93G^Br3zS=Z>c z2eEziAu*JckruF~+3h zeV>IFfB7fo7D|oIXOg9maqdr(cFq<`444*=7Pt#3_2` zdM=v(vSCd6E$qV}GpEUMCZ@quhfC;Hk7yQ2v@P$C&CreNn_&=!?--G_U2p}?NkOcidw5D;8+`j{d<#Hf>0U4Offl`+B zPgw0UMC-qB`5a1E`5fvO(7u2UET6*|ar-ANz%5ey-!$NY6y`a}KlJ?9+xTDl|IP|) zM&y5`4@=L8{6f?hqQ4OHh1f5|0n;<$DdJK7$Mo;<`$SkQCJ4xw0tkpe+D>{TO&(!b zGZ^mwNS^$kWdOvZO#nv(H3kxMB@{vbbLwHCg#Ri7q!muMCHsdaGEMkzhQjPQ3BEx1 z1tOq5Co!}6KPkY;WTSsC8b~~qOi%g`t$3Ae{x2QYb(Q>ulrN-yA?*w4U&sJ|py(hUh|NHi8I?t8*j|RKQWe59bjn3{pUZn|M=ZgOqi@MHzK>P;K2L^rSI*R=> zPWc?QfF~<*Cap8<+V zePw`yQT0^pXP8>Qa3YU9?BAqSc7*|wvxjrcQC9MxJr01Fd3K6n@!NZ?zVXSSP=?g{Y)7 zEQ*?PQVYJO#SgBW>3iw;SZv7!KU}i`)OFB@-i;U3T^-h64#gbT66$F?5TLGqTveSA z&cI+So1p>db#q`QY)fvTlH#btw)^$fl1^3>*QKCjMDN3u)o375`3}lgei8FQJLgFO zrvuG^!N$|VN&Z3ARO@5B58{(#0oO}U2fIG=zUK3OXQ3vFQfdW#MIXh;JmV_*GXZ}( zWY@ZRe6&lRoav63)T;m_tL|%WWDvHm-jslrU2Q;&`E@hK+Zw%KD&dwM8;M_MvNF9C z;$7EW%c^v^UFB-BF%fkteC<2em%KmrK^L9sF^x}vVyi%jv4UZeI%S+BXU{}DmJo~Hjjk6kz_};yF^dCr0Z%f`*3MyO8l|R{-JKzu< zgQo#`eor47x5KI}%@>~aqK{d=nOEQ6W<7zs-gmiU8!(5lYzrOAdlQPAAS4Z};@?-l z{WL4^88Mx7?uR{B6MmWM7|nv9MDLDgvw-k2{^@c$iQGw$Q6_s={p=*D>DGof&*>YN zH%CD+5$Vib)+lg&w{t7!xN_sjcc$&7^GYSXuIuIA}La)E7r0x-dfZ z!|SmS+*p6b^g9fcY}3NRFC;M7$+Wh6z>Yv4XN6Bj^*p zCzi!w@~Ygo<2ub%qUw<`znQ%Ola5D~P@j?muWa402DM4y(gdnP!R8`Ht}!_Tg-ZiM zFsb)FM|9sG?;@FG0%siR@X25VGC+~C-4?dYAj7hQ&-yBPr(Ds2>B7_#?qNz?`)KL`gO_u}BqQ>PAI2aUVp zVEhG>F4y=v)$=>7uS-nzxwX$nw%S^Wff%UkWLg`q|HL_O&o1;y)+;g_G`>?*8AJmm zLCoyh&po&SWhm~q<@S(FDmsg_4|tb4G1200mkVelrfd5Wio(yI+>NtN{1NmW-Tm;nP%F}z-O|eHPA=LnJ;z$%%wEFyzrRqS z+YN<5V=H2?CtdnkFpTHU+>Z^w0F3)58_v;X9G9$vw5qh!L`+CqwtN7fu=?KSFw&I_ zoWL`61CXt{j7_>>d|s?!8yvtjl>7(ckOxCnGVaK3mB1x5=$vMdBw?IGj0iQq7w6Ef zflp%%iW1Lg7)36JnhzrRry<$sNSoF?ez>V{0X}{WnI=sN_5Z5tPT-;H-UomWgF+~> zgsc&=MV73E?1U`Y3n4`aWu5FL`!-69ttdid$r2%Zlr4&^g=C48WclA`dV9tD`}g@g z-;a6bIp>~p=FXiNxp!#IT}kS9ThVfB)5cQ=>^&L$V?KfT?7f-3g5q`8OOhs=7tDKG zYpOm6O3=MnqT5>$?)=u1K7}ChE{5|q-|OSMCvQ{l(z})`bhWmyMfWP-e&fiC`A$*- z8KlOT6V|s*N;ON8KG{DXpLtMqqJzKcqHU^XMBar#YT_^o35Gk;w2fYKqYeZ;{!M)z zow3aQQjgCjcboO}8j$BFys(N-*(dygLz|-KqTQsUmx7vQIX_vgI?WC85V!lzm=|B% zxxSN%G)gAS5cQ;{6%%tVl%1P!oVY3)PS$jhAMIgdKP-1sk*7yBbH02W?J1Ex zt>MgCQKaTv&thjaSG7?J1RJLDU4bt5s#mA_Ugq9QEA{G2o7{ELCV>`v>TD9hFU-C? zf{LJz`Y1{_A6o?*=@^=>G+K5!fp{rnrs7klBz?Gr_9%n|5|Ge79yw{6yZ@2QmqwKs z^|h9L>=$ncah~!k*7UZxAEc9cD_oCfs97Smn(JN4_%(I1RR2(!E}8qC<%wAm&Tkia z`J9gx%c$EtO%qX)vh+7EIDU5{H*#7J!+mD%v;kY%>oXgkPZ=LI1-$o99wSt3x7Mjx z)_gz4BY)lhjfKPO^LIr*TATD8VYx=VSGQi1=OXRfuG;7ie91=Y%Z7u?Gi@X^$&X?p z?y1DJ^aTilxrgi|S&iG@@Exs%3XL_EmAd?I4{1YkcM1N!(g`w zlG^SgMfa?Igoq{Bu1@k_iQaYMF^NLxvU~_;YKp6`)3}E9%~Se|{>0WV zN(kh*r%8sK&np-Vs*`y7#NH?GIL_E{D3a|S+j$OA5ffV%rjKe;FQTR?7->wLC|D@T zpM2)xE!Ryg`pie^n=*Q{_sUUEvP+e<5!F^IXL&iz73>x~N(tx=QA)O2@*ZbpSE{Z3 zg2G(SF{5N?^E|&-pKD)Dlp)!p4_Yla;ic!+(1s! zJQzhiej#0b)G&B|gl^Z6qnO=1tqEm;nZ`#pRZR5ZS+Ya+2?NHf4tv9dcE?@M&Khd= zy+mMm*zw!pAmf(3aiZV#B*mI8eEr41Y|NKhe0_hod?rbpq$r>w+KlEZY1jmD6}DORv+m^49J z(x=D>b1P(2Ls7=^pq$@SNX-g;M_yB2<@5ZG-9dga!lwSs<1@COa~pCc`rbLF9d_wn zl4hy@#1}+#RezfCp`~rOx#@9*T|LGX=lrCzOO-ypZ>+dzXv^PZTAQ_KFQb;3{!`~j zXWU+~??g10>dLiBOV8I5?>}BIQNKngYI&tVg=QJ^-TS>CqtTJ(COX1b@7w2?XOh#> zN)x6PUS6GEX8Uv|cjAnrSLW39X^U|ekzm?aGCdDN+G(u~xILt<9;j9}zIfj2rDw~8 zsH#>?piF;JqievUEW0V2aWg9#<%NWJ3ku%!V~o-Cl~0@W-u2kdMYlRWD1K_zuA%&j zl2ar}M;KGmXP`#cb?G4X)a1%?wB^;v0;Ylsw>~fSRpd{1gr5&xpb;828{adxr|;?= zNhLa`v-As(+HmNH>m{1v!;UxnNXC}d4$gfZQm)!`$|S}$QVly+T`|DPKFIzq^cF&Gsis+cPuZjSey{vwL!H z?kM`eE+575gPUKHsR;&XYY7?#o60`#dP;oVJY=yH!y_+myzcX>xtD@G@8HqAr^lG& zoQ~U;eP6qKV4#aXiOYSMp0QeXC|HE=Se9e2I>|K|#$M91G3?cIWJELvrk-a$(yC%} za97E7OSHxsfXb$EJq*7thRcP}-Dr;`CQH4!8Y?@iO)py|lwEliOlu=xar^_{9kbzD(>$Mqvf0+Re(u;KWv|TXeF~g6h_P1~3X`96 zpLiCjGT+;C-b>~}TkF~L=fWi9!JNl-l{BAr-xQGi>_`>tz)+@CYsmUNpZ9Hs`K2zmHZ6&$ z&s~+$1B;J&hB7dZ8BILess)iE4lG5b;pg&0Z@JEX(6@cnX6oQ-`L3+-N5+}Zd*SUB z-_8njG3Iz%sqcy|NO~$=cG+nD!08^En{LA7_Z;}GqokjzHb{&zFA=%1aox?CBv37S zq$eLSNj9|5u^s zhq%|adZZ*(w56J^bu@PR`W=*=(Jg7~|7?ER>%HFFd%5vv4h5A@i(wu}IdCtj?IO8L z<`^-0UACG6b*L+jJWQ>Um!94&s&l_-+sl9(PbhEdCrstD_`Z(_@tWXFahu6vcGcs{ zPjvGq=v7L)sWF5(sG~BI(AuKA641@MYqCJw()DT^` zW9JpfgtGYA3~tsfGE$cuK58Wscj@Dapvo(XDt9hR9=`O#U3R)|l9ANk)ks4r@iF1z z^r*A`QTi#gyieI^xRJ&U{)6qK0kt9bOGjEisr&0aH?N4fP=*=2%**rGb9ni{^TF>% zw~$ZW;XRKfI@ym!B$smfJ$rR`if@>Xi;d0FGmtwz$0cg9fqwWw(E*~50(#AL0v^rm zl;q;7RD+`%eQd%9bzcdd84|FXZqZ*^V4AzD zSmh1PDTzBq4k`HqnB4X239VwUWXXN!RlC0mcbrMSJxL!s_(WCzJesn?=`+{)qj`Z= z6qg@f6fA14_D!i6(0=L6)J6HF)WOXDOM}uJ_F1I^^HYaHPyI zES9R&KEG0EG;(b9E!Vj=FV5++Rl68@g*@+1r&gxDy?8>cSL4vWAg)C7$$oCy)hivo z6JOe{D$^3a_-Jp|Wsy$%y;o52n{$kA`OlAC2bdEnn%buB%VL)==U@gHsu%st0+(GE zHxHCfpOzjXnd@~cx8Pi!?J(pYxODto@3n4wmzv%e`Q_M7(U+KaFBmtf$68jk7R#hf z9Lp@*20mCRz7h;EP07C^eJK%>*!R}KX@1Uo{P=^s#;XF~bb>BjwwlxkPo&j#QuJrH zlgT+-xPH=>IsHgUiE7&zrg;yqx9{+pF59W)FuuJ-X3C|;Erd)9m5Yuqlc$#q^Tos4 z$V+o`2~1pXzuf)ttK>c-%P_k=S0k=G+3&06qy4;>e*C~ou4IlMZP!1FP7tlys4fWQCynVkFe@iFJTKlUco7u`O}`nW1_k&TP_tS{GhOwJd*FLp-? z?u%$1AnntT!uAM~TimA4F?7l{#vC#_#N8F;%iA@0nod=tRr{dzhj(t?F+~DGE4OlT zJcAeyTP-)rWYBRvNHDXg)ueIKP<1l6Vj^mP=w4TQb+qGX@UzUt44S+GXOjcyYNJvz zqpDAGRBe~|h_kpmyJ}O@mspExj;-jvz%VkX_L$Q; z@>CaDb!F~z$TxpS*hkXX5qL+>@zI%faZ0s}aqH3N$L|#RTCG+{?kXeT9MAu1Q`~Zv z49l?Dw0nYu&SC8m_ZgxwQ=!*3la6&7Y!#z{OTc<^uPg4b# zG|4w3wcBu7^uy~BU84B>$%2G;5xAQ-6tDQa;2v)yuSyp_?EY)-Jtv*h5ty)*5NE~s zhby1LyJQNEHdRe*GI`Z2_FtHG*>u8=CFj(;=tW3o*pCM#>+2>6Yb$6S(Vhy{a(@2; zcfSx*LLk;Es@k$IHz{exHuai9&GNbTTm)fvLJowa&QJG?Uh6%~{NB!bDeKY$(eTuC zeXF^!MRy_ZgE!u#JUdbt@$HfchM3aSLw07*iRhRR!z|&iL1j|>`{hi;o-8)m`z(`H z@@2IgP5cs3GC@Qm@Q%UK)=KF6muu^pYEAyp-A%?8tLh9n$3Kz%Qi_PFiu%}E++iWT zQQenuT5|tFVJu;k<`lCntamc)7^Gy_X=9qgwAp0 zt(=*rdL-mgPwf`WF!?~VN<<$mY!`mKotjZG#f!%2;6cjJ2Sd5DWVTGt)Z@s@YxZ4u zf0av=OtIG>I3*0jQn6pvIr9BkQH+sP6)|VQ!|wDmyWULGE56sS>qwf^4mW%KlY{U@ zSI!Mlv-tP>y`Rw%mz<#GewiVPxOL|F*g3|rIx%H zz4h{}jA+4}N%a(4NdwcSQ!_j{yF-88Kb&ABtkFB-i*xAye0?fl^|pGprd5$}wZ@i_oN94!g>Rgv zX(ESHk=Ju8jJCirnuk=s8e1FqD{*BTrEi+~<)i$N}n~P45Rl`BcBMb=v z6-GJ1mzT2$?vbX?J9^6o`H3b7y#C&r`sfrb(pMN{K-J((BipUCXD ze6e`>gnnkrW!FO&B;@T%TXo27y`OKAe0(gOq^Z34YI${S<$9tyrjOq%{PRlw9kth) z9onw!lcfiro+tFA(f!f2)>HJ`-A=#<-3t0Vl#?=%(oas>^qwvlw!$J*v28swJ z&ft<$^8N+NJ%S}UpQXkIic{)bFO9@CK3n&S&b-$9qvJ=#)2tHLnmP9mB}|=f>-Wst z#9pZSDtP#!Rl+q970$`Lh>ed7(|iXVPD@@_P1DdhJ@ky~5;pm^x7oFaIAequpl^^o59$(t|xpX90#Cdq4wIw{Lj((qmkNUS~< zNpjUZI*jkS=YYacg^J_5HaCm+0dC~yigL-Fril+6LRUBu<{t6mkg0VElRw{;EX-7z zbuDOlE?Ue>SOPVH{upxeftewN|5K6S7jvJS-?1ZS7PN(~@AkcIdjDu)oOk^AKwS|x zpQdK*Et+^&n)9zJp9-hd`nY8uKrVbnDp)9R*Lzp4vxCczWBT9xNy2inaB_}d;i+lOJMHG zzo*gVT_}dOE>Ao@q?Crxmg>;=cl{HjGs2jSndXr=vJVeAuip1mx#9BD$=ZGI^Y_=O zG*^cjH!35ag$t;lKs3Y?=w4;AH_bMufiU{ zm`!Cri>k~J9?q69*5{e)*c(~w>}AjrM-V{LrB3}Zj*cr=ZB|sDkb@0<@w>lKX95lR z(4^$Nexe`dbF}QZl+tGzG%1}ib?Eqmvp5oIs&q>Zt=7T5A-kp{byL0$k|B!s4l^J5 zR5)?LCIV|z?DlIAms~OSv5VfzVX-o|De#`G@&{{mHLJKm<`-7UmN&u)jeL%cKhF%2 zokXXt8WWWdmp{R|#&a?+f8y@p_0IjmNphhmx>kYgAx7B3R&&!yMmXOVts1RU8TFZ` zw0^f|d4!EAYiDP`vhD;W&xqg~Pj;D$E{|ht>prtMM87J0H}Es_ZavG{M5+D(W3MAm zSCjm0=o&4$r8a60Jiug9-ZB9|0C0FJwe%vel{5roL=dR+jPe=76tY6B-^6ii4JsDjf zQxYvU!(6v$#;5)Rh6=mL((2E3=MA#vk47@fdV7TJ|24%g>HG83*z)Nn%^LcH zzAuFZx#`}&?2J(yrO`i%vTb{G`_1zn8?RIK{p87p4ihXG{nM;QL~g>jNo2f-Zaj44 zSB!LSz3t#z@W}txF2AD*W^S&~NN8@Kno?R(=9?0Q!UVuB(!zkplT9p03ie>NX zT`P;NsWXJ$GS_eQGX?cDyj#~>>LzAu;4`62H8`NZ{IZ+Soj2)(n;P=XYc$U9dBUw# zmt+)9-n|sk^CA*6z}6ot&S%@~(sWU3a5hx&t-@K~X_d%>;z7Mn0u1bUYrN|2-@4<~ z{Ihm;EI|0ieTLV&&gI^?TtjnD1KZ$(LcWxq(3j*y|6#R)wUv6xgoc~6bbH(MRjmYL z*cnc^t|o*MCeHxTzhCP@5ugI^~|(z zWAjP+0*_ab`yRLL=6R}6oJoz2m|RX!dCKn;(kG(ly{G%t39mf4s5h7Qi!MaQb_qqF z;mq{2Jud1@@I+RJ-hjZ}JA#Ad$3ug1*v=-n>twV{S7wYN0o>XPfG0upcAPwemO=+i$| zPMy;G;eDzL=W_Xc5HH58y{b`lTxrbLo#&Z!$kSSq15>$+noE6iV%!!Z%g?{uf6m|Q z&u@Ue+PeG0rSi35t_XL7_g%Edq}n4g7|b(2GhSU^{M=D7^y-R`}v;^DDPx_NXaPu44A-sf*pOSmsfk z@-djrd7FMDH;d0%!oBrK%1u(|gp4i|PtF>DBFV;MM_HIVkBAafErwSeRsJ>d^fkV){2dPyy>nfa2A3UK?WfpHRo~Qph z@s3%LmQtP=?cZQ&&o(Gu;W-?SvG}lCl!06Qa*g8J<$|TDmMi41zmXH^Zuoh2aO=!F z-^+Ry`NFuLrSFi65^1H;&Fc?*Yu^#Q;E}Cl-ldIgu0wVjRa@eOqvJWLa6I7l#E0w zx=*ZA4u+G*k9aG-$R-X=vu$+=%4iWavKD6#xVi7+aTAt^Ri6RrNq^k*5XOVf)cr&U zflEsEcL~qL?(m}udVEbm9Mi9FUX~CP<3%&q>vPubcC(}xja}0XW@qA-mpk)rQJcM7 zjq%oM25knztLN#FcNL`Crlb5dP_vB=Z{e~jite_V$`E-Lu1cpDh8*uzA+CB&s|un z>w3>>82!QRh)(2;>0QsZJ;p(j+FEbER(+X{pdOJ?r@qMV;d-}5Q+Z!~sr^se``(7y zRZ+LmXL)s_#B_ds)=SG=3wm+lf}7e;3=`=RcWC5eB4rMXRyvVy7o^U`a1-SOP5TZi zH0SlU*b}Z98S#vtTxje$Xc{Tdc^Vr(70X`Xaq~29&+^*0)booKv0mj1lDyof&)$oq zd35&Smp(JQ)}`*#3mMU-KZP#FU)GI^=(9~Y7d!h_@71f>*5*I%_{&9o(XrM~^Sf^K73eSs1s&6Bd$^K%+RHG|Vop+O;q1>1v&c9n zwa2-GM@I?sUEZ$yxW_S{hq=mQo_# zf$UQ<^EwlDMr9(DZ!UcdEDV{YGwb)n>gt6B$L*WtkY%MjDdnnh=+VA$c9K0=3ywy+ zu3;SSI5@3Uf4lJ^t44e|+xM)ja6r>9@5duwt=GQJe)u&@b|*JvZ9Ize-lqb-YZDZd z%OYbat-Y^_zHA2c5}PvAU%PSC$Hw8>Dc`lPXrDGJ78286CruC2y6-bC3mobiPB@jg zQAhUDnMO%H+F5S#!bJBmcF`}i@(RWNZVeI?Sj=07rM8p(RzHL<=-E)ljXVfEX;xU~ zW@wr;(O7v1onCk0l<2`b54xwXJG1dIM14}?$}3fNh&*GAw#u)M21#Xzy@UVv0o36Je$6-()Ca_A-l7 zxabsMr20QTIJKq~{`xMZ^a)X!aHdrE^Ap|Dt8CP)7Nv0)cM;A0cp?7`OSVxFFC=rN z@jPqf&jr_~Kff7G-fye^QR6sv!jFY&Gf+ZZNNFB(2`j&n?aL52{%IkPcj^jx!-Bx% z!`bf};+n*sfP)&ujjO?>`sh{-y+w?uD`EypE2`e*<}U8b?FB~hzo!E>DMCfkBjXI^|>#jcagsG!!4tV zu(!H$q8^ymMbS+AJ))7_9ZU9m>Q?)xed?OUIr`Rx+b(hH9y!)NuUCuwO@HW}NoDH? zi|n&v{dMaHYi&yF;(m)4AS1^KSP4Mo@#kZQhU%Lx*`5BW^X&eCN+XsnAevt;$pJmNQt?zQ;UIG8@v|3%x5euLNjvg?(m{G-(ILR!aC z4zFME@9)zb?bsl$&-;OmKSJt97kJcIz;@&5&8v?vQ5S=IVuV{Y5-y6 z+K|8&%zw3jLKBl&B-TSGD$Mmu(^Jm0p64))pqToheSsJQa;+q;k4ep16{jlrhl$ukWq`TUZqH0J`VwF`?C zllNbfx>3{=U!tNjZG7AE^t%-2cuq_tF05YricGJe#}iiFF{(!c_pJDYkG?#j!=P|Z zEhLX^MI9zL{1Y1q}f9+{ca$!a;s4NvMO(<(J7+HV$W;+Mp zr%RTV!yZrWv~wrK^)H3jTYFME6%-GDEB6e~Vb3=t>CdZKm_FyTbaE&Svrc}`cjaq- z$Hw7DWgAk}KZ}CK^Aj;K4}|J$HfjtK{f;i~s{JuO{2}x7aP#!0mv5x6jzOX&Tb$*l z++c|iS=38bS{BUQr90p5eCbB69q=o^%;D!A53+?OFR{p{Cl4?hd*gm3C_+S28&)XyrhY-uyfM<(!%^ z#jn?rH^{UfWSp3FXBlOfetKK{Xz}Z{#Gnj1pLIg{>+IYmy&f$tU&@>JGp6J-_4u_&@r4&*rWG{NSMPFt`81Vnz4&QpCd5gr-9bHbeW^^W zTt!g$b<6hvSL=|0@~Gu=X430MI<3JRrC(zCUU+u%tA{<#81oY<`gwAlhlQxg$YFs1 zy`Z*Q-6gQOTj+CV{Z+va`gE?wRX^HaEXkY}J8{1^n&Wxi! z7kO|2_pI(PW+riKuL}|Xlsa+Byh`Y%giy0IE4QQe?UA4j#TT5OBgskgpQeTy)e_kY z-QQlT@~)w)A>eAheOUwJxLG>qbpMyeXGdb8ht%5X4`!w`bWNJNSe{+!{u0<7@M2Ns z>;7LKyl)xEzX>HPILFDBM-ssDQS#HtLn1gd#X~=?ibBqpCrcR?S@s+icTay`L2LH# z^|xEOWzsLN2=lOrQmkqGpz9u*C+c*UT}(;oo?!f>QGC2l`81;~a$y~eenH?{b;mQ^ zhjjIanU4$v9eC$oEjSmNa)ONQJdx^+`~JVqb8oCo??2`EKKw-M;XPx!8JapeI&DuE zo@nhMOT8axd^MZg{BEo?)55d(C{798I)N0T8y9kicx?C+LR;(=CA;%(DvzQx-s@>g z$u^XTaB9zIo%`g|hf^Z%Y^}x6d#Sij@P3sZ%byu_j_gkoXo)6dQI|v+@RwT+^%5<7Aq%zSS+OYT>$2MqoY_wFurww(nTC48|_FEx^ z-`z;jy!Z=!^%w0sD$zfP-49sotR(O&V@)BF%*iZhjx7fH-#{$m-QX{&1Cta#7PeOmR^=Z*+ij`vrj~r0q#2aym2n@+`BQPV;24k1x4t zd*WvOHd`d3Lh29Kuu`f-{ZNkDTv(yP+_0=U==-a1(^-D7hS+yK)I(ZW>~Y}uk`czw zceOr>9_2p2?0fh0h-Q`L>`G-Y^Nhi+&99pQq25%iC8JFel$VH)KcQhh{|ciZm`hnp zB2KE~^_gm=T1Hw{&hzJ|iS9!BOF>^+cMn>6?A8ezs}dPXHtO8yk-JU8*JAK1K9@vE z^<9+Tb@x8Mo&oYml!XUg^kTYPYMYqCRT)BfaYRh@Jh;Zu zs@q5JUY`&nP|`LuI@@G-W7hJ%GUJ(Vx49lE*cVD>9WkZIqmrG6O8gY(+vJi?|z*YO^WX=c_3};G@8wUG-~`AO>s`bOjx1S zG}b>Y@ZD3v=^M5c6>0nuC{&sd0ggZyIRgdPUQ>dW`1@p?fz>o+3H+%mMNI*<+TY;EytVYf2#~h9)uc_WFwJ-#6u(= zA(4wj9ukj{$VZ|8i9#fvAW?)wF%nOaC_$nWi83V0k$8qg1rn7=R3TA~#B(HSkf=rC zMVTQ<9l;I!%uGV+{+)1WMMzy)MM(X%-)E~G%dX#=M_P7$Xj*my5-*WxM4}0aW+YmW zXhot8iFPDjA<=UWyJ3Q z{ys6u5$!+C%fcl8q{;QRnv~h-jocK+SI^%Z**Ztg@YaYeEq}@$BoF56HJwzc{GwLi0=m3Be}2#21fS%soB~|vI(e`*)5wygwB@z?WXtLxkJqg#9YEd8h$!@q;U6}Fe4 zDgS3HjPU5cJ3`s8@MyINz7@`TRPoO^w>CjyGiviETjpXj8bO9HCb;s~maVUhDsbV` zIq{=vf2u=Q;zyAfLt-3>2_!xu@d=4ZB&Lv4o#4jY)kXT1z1BuPsEAa$TNC1QY8Xy9Q0TN&rup1x+$N+L+ z4?qD>0#v|WU>`sY&;a`ZT7V9q2N-|@03*NzFarkx7JwCC1K0r$fDtX1wau{0!{(S zfC``rr~&GL2A~OO0ouT6KnKtT^nf#fK41VC0!F}Dz!)$AOaU{%9IyZ^fpfrlzzVPi zY=8@ZEno-O0~Y}Y;1X~da0HxyD}XcL0=NQK0XM)M@Bpp>*8xw!3-AVf0AIilzyLP@ ze;@z|1cHEIAOr{nZUVP}FyJ;24nzQvKok%S!~n5C91ssA0Eqw=NCNHvcY$Q!9*_d0 z0yrQINCz^2Odt!m4?F;}fgIo=@Ce8S@_@%cK2QJ@0#ASe}u>n1OOp`28aM+fCShD>;^~yGJqV|15f~z02Qzo z*auJpG{Am<7N7&@0S4dzzz8q_%)mi_1z-i(0Cs=_-~*bS08!u=AO?s75`ZLd9FPK}0U1CRkOSm_6TnG80Z;^#fKz}npaQ4@YJfVR z0cZkRfHrU%&;fJ-J>U$W4;TQ3fDv#OFa}HjQ@{)`2P^vcnkCc1He1rJunCi z0Uv;2U<4Qi#(;5P0{94g0w#eeU>f)g%m812Szr#B2Nr;@z#^~&d;^w&@4ydW1^5Z9 z0>6MYU>(>1HVKi#9~$|A03Za=01-e8kN~@Y-2f>-29N`L01AK-paS*+`v7Wy2G|eK z0(1a9zyKTo7y%}L88`^A0IUETzz%Q#oWLP~3*ZJ013UmPzy}-w_yGYx5D)@{fun#3 zAPO72f2XGCz4tN4yfH&X+_yT?a2DkzE0|7uF5CjARAwVc_6SxJ0 z0k?s0AOeU4qJU^128adXfOsGQNCdDz5^x8&3nT;gfD|AVzyWDMI*=Y7f;QgKwgbu8?Z_Dgm!R$OxvgazMRIC8>NG>PEd+kxaP=PCFmi|4j$(7< z8a_KGGmYeSu7=0W67-cFZFk)&LA&k9r69ST3vn>)&t$d+{s75|p{0@iVk@_fvC z?i!LiyQ9CaNKS7@?n3l`>xPls&M_Otl%P%UZMW9yU^P59W|4dNcl0V93(t-1D_f7` z40qPWm7v{s2WOgc7v$j_nHKpTRxIJpwy=v8w)Q*!yCrC=9d*RX z{~gP1BzI*;9rL|ExvkZKCn^)J8M#Mrs~Z2*TuZ-dhdY>3f;QT^1AkmdL~?%(AAeb< zB<*a^MI2WuJdyu>h}qyb{(t?B62*>U_~+(c)a=O$avXbR37T`afj%*6Ya72WD#dK_ z1xJM(h3N=s_xvuysUtTJZS{^HOEi*WBIFYMr{pzqlXj~s`~dFR2b2ZVqfp-GP$=Dh z20(xeBg;SfpJv$Cg03PYe~$s*cWN5+ou>S|Z|sP&tJHssWYeKY#N(eLq^G^O2qee4 zV-&9MikRgXQK(-RQ7DaVqiFX2yDdw`wzl|zSR%Q=9TO))R?c{98~8d5k_*_8A*BG0 z41T}`A@bo%w)V2ua1Z*f=2dK`q8r`+dccWgW@m81*~kNqp8j2S80U=KWZ1E>M@Wtd zo$vCyCv2L_RpxCE3Oh^&zUGcj7=?OQgF=aKTWm)fHyDBl?n+Jxnry3Y{GPJ;@jGiM z8igv{g+kG6Yqq3_D|m?fZ`F=z%^*1@bd!o3HpU~2_V={lEydQv@lP)_tt!5P8&~tJ z1Wml-m8tL%bmI1>#5cDDO^T(obfbrPZGVGu9J!APo%`G!w#SWwQr@i+{M4n?lGC3N zqEKtdJf!}a13c%l{FN$^uk}KjVLLMkxA%>2Nak0hH8)3uQYRuoo!HjU<@xW1xO0Wb z=o7OE8U8!6jyxFIz+>;QZ z&A$eOA2TyDVh<2o9z>{Q6fsI^+q(Ul3jW$1jzVcgBLB;|ZI9zV7Y2=0|19;^F7@~cjOkhc{(dIhji)8(JzKlp)+@z< zBB*=2tqvC*vh9Jv&tTgX6t@lH0!<7I;-8J+ud8@)YXGx8+;?4Vf~Z zhATpzK7&GG&k~@HZEHO|u+^GlTSNRRlOcykCiFc49DeEW6Tu>zAjN4RckIFMd0XrJ z`_RfK|4%V)61Q*X@fAOp;aTOY>@g@5uAKlS{qJhS5%JcxwtN+^@qZ6A__p|~O)8*J zdxZ&5YTIVQ`etjp5`SibAKZW5^&I|pqwo+H(co@WBFF6=Pg#7c(@$Xwj`^_;pS}a!iEp&3>0) zuONRNvTX|ZBK$jfwEB}Y?7l0bbU2;o$o~=JERnbOJ9kOyFGp-yk(bQlv?!F$wn1cX zZ`_PD7TGZrW~5stw9d1?wj#8nBpqKOU;kGLCAM=Qjr+f2#=oyemyiEb3V&NdwPT_2 zkLzE5hJsSxHe|YqzuHP|jRe2k+x{SBTeC(c-1&Os#R0olE`2ZV67p9n99wPi%UOWr zm+WR|MazWI6Phll`k*< z?hAJvxyim`OqKW&-SFQf&>7dZe(;l?qm$jtKrSFltb{_TZkzN8!e+F}-0eO@2zovH*xr2!i^LO2`twG}-JJc(uCzBXv)O?~A>iD+tWX<2l z#WW(X;y9L<@Ht@nbNP0!j-WXk0cs-wjZ)m!ZsSVbU+wl_(?aU@Zd(WZ%6~xmWJ32x z*Z;1?I%U`MY^%brR?KI|;yN4}wGl|ZW3~2w!bKy4`@LFQ4--KXbbB^r5Z~>wJ%ZL( zJ@6lh{vADj>wZv1II{~EgX|)7iLL2-aMeg^$0M5B3{?Z{e^tpNshuwbe)zHi>EC5o zVZ9N)Z8L!_rnW#=ajsikVcBkt@NcWb&-kxDe?duZTUYCwTN~Jm{gygH_owyNQT+EF zpM*RB$h>y$@%aAvW-8k)SWqZ|XcS6$TmOr{TWM`Q(D;G3!N6aS;(I)fGemCw-W9es z8UZEfKGZ1wbq4=OGkdo>!S7{H-Y_~bA)i*+O?JLB6C(KC3|mM)D*FFS2;FjE6b_s0 zJN9w>gkJkt4`+&?P|Itm9Xlc?W9q9Cv^st$Tcf><Br)lEh;XtP-*MhQi;G4YXl^yY?|2(XjtQ;l_j~I& z*-rRO_V2!lU#nLC8f+)hr|N$mp~}stIQG}bZ%}a-$Zv6}w+$OVoE)>G*uDFaH^9h? z?DpZ1apTyKmOF-2F;Lz|-AAqK Np-{YbCCGc5{{x-$*gF6K diff --git a/site/notebooks/EXECUTED/model_development/1-set_up_validmind.ipynb b/site/notebooks/EXECUTED/model_development/1-set_up_validmind.ipynb index f82f57eaa1..4244924b9a 100644 --- a/site/notebooks/EXECUTED/model_development/1-set_up_validmind.ipynb +++ b/site/notebooks/EXECUTED/model_development/1-set_up_validmind.ipynb @@ -171,7 +171,7 @@ "\n", "

Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/EXECUTED/model_validation/1-set_up_validmind_for_validation.ipynb b/site/notebooks/EXECUTED/model_validation/1-set_up_validmind_for_validation.ipynb index c5dc1fb394..05ad11c2c3 100644 --- a/site/notebooks/EXECUTED/model_validation/1-set_up_validmind_for_validation.ipynb +++ b/site/notebooks/EXECUTED/model_validation/1-set_up_validmind_for_validation.ipynb @@ -261,7 +261,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/how_to/tests/explore_tests/explore_test_suites.ipynb b/site/notebooks/how_to/tests/explore_tests/explore_test_suites.ipynb index 6eb1e5ef79..976eaedefe 100644 --- a/site/notebooks/how_to/tests/explore_tests/explore_test_suites.ipynb +++ b/site/notebooks/how_to/tests/explore_tests/explore_test_suites.ipynb @@ -137,7 +137,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/how_to/tests/explore_tests/explore_tests.ipynb b/site/notebooks/how_to/tests/explore_tests/explore_tests.ipynb index 56c58b62cf..3bfda30329 100644 --- a/site/notebooks/how_to/tests/explore_tests/explore_tests.ipynb +++ b/site/notebooks/how_to/tests/explore_tests/explore_tests.ipynb @@ -107,7 +107,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/how_to/tests/run_tests/1_run_dataset_based_tests.ipynb b/site/notebooks/how_to/tests/run_tests/1_run_dataset_based_tests.ipynb index 9af05b3b3e..c4937af216 100644 --- a/site/notebooks/how_to/tests/run_tests/1_run_dataset_based_tests.ipynb +++ b/site/notebooks/how_to/tests/run_tests/1_run_dataset_based_tests.ipynb @@ -153,7 +153,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/how_to/tests/run_tests/2_run_comparison_tests.ipynb b/site/notebooks/how_to/tests/run_tests/2_run_comparison_tests.ipynb index 1ba4627bc8..ffcd999fc2 100644 --- a/site/notebooks/how_to/tests/run_tests/2_run_comparison_tests.ipynb +++ b/site/notebooks/how_to/tests/run_tests/2_run_comparison_tests.ipynb @@ -1,1094 +1,1095 @@ { - "cells": [ - { - "cell_type": "markdown", - "id": "1d29276f", - "metadata": {}, - "source": [ - "# Run comparison tests\n", - "\n", - "Learn how to use the ValidMind Library to run comparison tests that take any datasets or models as inputs. Identify comparison tests to run, initialize ValidMind dataset and model objects in preparation for passing them to tests, and then run tests — generating outputs automatically logged to your model's documentation in the ValidMind Platform.\n", - "\n", - "
We recommend that you first complete our introductory notebook on running tests.\n", - "

\n", - "Run dataset-based tests
" - ] - }, - { - "cell_type": "markdown", - "id": "61065444", - "metadata": {}, - "source": [ - "::: {.content-hidden when-format=\"html\"}\n", - "## Contents \n", - "- [About ValidMind](#toc1__) \n", - " - [Before you begin](#toc1_1__) \n", - " - [New to ValidMind?](#toc1_2__) \n", - " - [Key concepts](#toc1_3__) \n", - "- [Setting up](#toc2__) \n", - " - [Install the ValidMind Library](#toc2_1__) \n", - " - [Initialize the ValidMind Library](#toc2_2__) \n", - " - [Register sample model](#toc2_2_1__) \n", - " - [Apply documentation template](#toc2_2_2__) \n", - " - [Get your code snippet](#toc2_2_3__) \n", - " - [Preview the documentation template](#toc2_3__) \n", - " - [Initialize the Python environment](#toc2_4__) \n", - "- [Explore a ValidMind test](#toc3__) \n", - "- [Working with ValidMind datasets](#toc4__) \n", - " - [Import the sample dataset](#toc4_1__) \n", - " - [Split the dataset](#toc4_2__) \n", - " - [Initialize the ValidMind dataset](#toc4_3__) \n", - "- [Working with ValidMind models](#toc5__) \n", - " - [Train a sample model](#toc5_1__) \n", - " - [Initialize the ValidMind model](#toc5_2__) \n", - " - [Assign predictions](#toc5_3__) \n", - "- [Running ValidMind tests](#toc6__) \n", - " - [Run classifier performance test with one model](#toc6_1__) \n", - " - [Run comparison tests](#toc6_2__) \n", - " - [Run classifier performance test with multiple models](#toc6_2_1__) \n", - " - [Run classifier performance test with multiple parameter values](#toc6_2_2__) \n", - " - [Run comparison test with multiple datasets](#toc6_2_3__) \n", - "- [Work with test results](#toc7__) \n", - "- [Next steps](#toc8__) \n", - " - [Discover more learning resources](#toc8_1__) \n", - "- [Upgrade ValidMind](#toc9__) \n", - "\n", - ":::\n", - "\n", - "" - ] - }, - { - "cell_type": "markdown", - "id": "67a4d9dc", - "metadata": {}, - "source": [ - "\n", - "\n", - "## About ValidMind\n", - "\n", - "ValidMind is a suite of tools for managing model risk, including risk associated with AI and statistical models. \n", - "\n", - "You use the ValidMind Library to automate documentation and validation tests, and then use the ValidMind Platform to collaborate on model documentation. Together, these products simplify model risk management, facilitate compliance with regulations and institutional standards, and enhance collaboration between yourself and model validators." - ] - }, - { - "cell_type": "markdown", - "id": "eeb30df8", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Before you begin\n", - "\n", - "This notebook assumes you have basic familiarity with Python, including an understanding of how functions work. If you are new to Python, you can still run the notebook but we recommend further familiarizing yourself with the language. \n", - "\n", - "If you encounter errors due to missing modules in your Python environment, install the modules with `pip install`, and then re-run the notebook. For more help, refer to [Installing Python Modules](https://docs.python.org/3/installing/index.html)." - ] - }, - { - "cell_type": "markdown", - "id": "293c3f98", - "metadata": {}, - "source": [ - "\n", - "\n", - "### New to ValidMind?\n", - "\n", - "If you haven't already seen our documentation on the [ValidMind Library](https://docs.validmind.ai/developer/validmind-library.html), we recommend you begin by exploring the available resources in this section. There, you can learn more about documenting models and running tests, as well as find code samples and our Python Library API reference.\n", - "\n", - "
For access to all features available in this notebook, you'll need access to a ValidMind account.\n", - "

\n", - "Register with ValidMind
" - ] - }, - { - "cell_type": "markdown", - "id": "4fc836d0", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Key concepts\n", - "\n", - "**Model documentation**: A structured and detailed record pertaining to a model, encompassing key components such as its underlying assumptions, methodologies, data sources, inputs, performance metrics, evaluations, limitations, and intended uses. It serves to ensure transparency, adherence to regulatory requirements, and a clear understanding of potential risks associated with the model’s application.\n", - "\n", - "**Documentation template**: Functions as a test suite and lays out the structure of model documentation, segmented into various sections and sub-sections. Documentation templates define the structure of your model documentation, specifying the tests that should be run, and how the results should be displayed.\n", - "\n", - "**Tests**: A function contained in the ValidMind Library, designed to run a specific quantitative test on the dataset or model. Tests are the building blocks of ValidMind, used to evaluate and document models and datasets, and can be run individually or as part of a suite defined by your model documentation template.\n", - "\n", - "**Metrics**: A subset of tests that do not have thresholds. In the context of this notebook, metrics and tests can be thought of as interchangeable concepts.\n", - "\n", - "**Custom metrics**: Custom metrics are functions that you define to evaluate your model or dataset. These functions can be registered with the ValidMind Library to be used in the ValidMind Platform.\n", - "\n", - "**Inputs**: Objects to be evaluated and documented in the ValidMind Library. They can be any of the following:\n", - "\n", - " - **model**: A single model that has been initialized in ValidMind with [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model).\n", - " - **dataset**: Single dataset that has been initialized in ValidMind with [`vm.init_dataset()`](https://docs.validmind.ai/validmind/validmind.html#init_dataset).\n", - " - **models**: A list of ValidMind models - usually this is used when you want to compare multiple models in your custom metric.\n", - " - **datasets**: A list of ValidMind datasets - usually this is used when you want to compare multiple datasets in your custom metric. (Learn more: [Run tests with multiple datasets](https://docs.validmind.ai/notebooks/how_to/tests/run_tests/configure_tests/run_tests_that_require_multiple_datasets.html))\n", - "\n", - "**Parameters**: Additional arguments that can be passed when running a ValidMind test, used to pass additional information to a metric, customize its behavior, or provide additional context.\n", - "\n", - "**Outputs**: Custom metrics can return elements like tables or plots. Tables may be a list of dictionaries (each representing a row) or a pandas DataFrame. Plots may be matplotlib or plotly figures.\n", - "\n", - "**Test suites**: Collections of tests designed to run together to automate and generate model documentation end-to-end for specific use-cases.\n", - "\n", - "Example: the [`classifier_full_suite`](https://docs.validmind.ai/validmind/validmind/test_suites/classifier.html#ClassifierFullSuite) test suite runs tests from the [`tabular_dataset`](https://docs.validmind.ai/validmind/validmind/test_suites/tabular_datasets.html) and [`classifier`](https://docs.validmind.ai/validmind/validmind/test_suites/classifier.html) test suites to fully document the data and model sections for binary classification model use-cases." - ] - }, - { - "cell_type": "markdown", - "id": "8d52b6e0", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Setting up" - ] - }, - { - "cell_type": "markdown", - "id": "e0d2daaf", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Install the ValidMind Library\n", - "\n", - "
Recommended Python versions\n", - "

\n", - "Python 3.8 <= x <= 3.11
\n", - "\n", - "To install the library:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "fc97888f", - "metadata": {}, - "outputs": [], - "source": [ - "%pip install -q validmind" - ] - }, - { - "cell_type": "markdown", - "id": "1ff56571", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Initialize the ValidMind Library" - ] - }, - { - "cell_type": "markdown", - "id": "c4d9f164", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Register sample model\n", - "\n", - "Let's first register a sample model for use with this notebook.\n", - "\n", - "1. In a browser, [log in to ValidMind](https://docs.validmind.ai/guide/configuration/log-in-to-validmind.html).\n", - "\n", - "2. In the left sidebar, navigate to **Inventory** and click **+ Register Model**.\n", - "\n", - "3. Enter the model details and click **Next >** to continue to assignment of model stakeholders. ([Need more help?](https://docs.validmind.ai/guide/model-inventory/register-models-in-inventory.html))\n", - "\n", - "4. Select your own name under the **MODEL OWNER** drop-down.\n", - "\n", - "5. Click **Register Model** to add the model to your inventory." - ] - }, - { - "cell_type": "markdown", - "id": "852392e5", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Apply documentation template\n", - "\n", - "Once you've registered your model, let's select a documentation template. A template predefines sections for your model documentation and provides a general outline to follow, making the documentation process much easier.\n", - "\n", - "1. In the left sidebar that appears for your model, click **Documents** and select **Development**.\n", - "\n", - "2. Under **TEMPLATE**, select `Binary classification`.\n", - "\n", - "3. Click **Use Template** to apply the template." - ] - }, - { - "cell_type": "markdown", - "id": "6490e991", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Get your code snippet\n", - "\n", - "Initialize the ValidMind Library with the *code snippet* unique to each model per document, ensuring your test results are uploaded to the correct model and automatically populated in the right document in the ValidMind Platform when you run this notebook.\n", - "\n", - "1. On the left sidebar that appears for your model, select **Getting Started** and select `Development` from the **DOCUMENT** drop-down menu.\n", - "2. Click **Copy snippet to clipboard**.\n", - "3. Next, [load your model identifier credentials from an `.env` file](https://docs.validmind.ai/developer/model-documentation/store-credentials-in-env-file.html) or replace the placeholder with your own code snippet::" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "c51ae01c", - "metadata": {}, - "outputs": [], - "source": [ - "# Load your model identifier credentials from an `.env` file\n", - "\n", - "%load_ext dotenv\n", - "%dotenv .env\n", - "\n", - "# Or replace with your code snippet\n", - "\n", - "import validmind as vm\n", - "\n", - "vm.init(\n", - " # api_host=\"...\",\n", - " # api_key=\"...\",\n", - " # api_secret=\"...\",\n", - " # model=\"...\",\n", - " document=\"documentation\",\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "99e9d14f", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Preview the documentation template\n", - "\n", - "Let's verify that you have connected the ValidMind Library to the ValidMind Platform and that the appropriate *template* is selected for your model.\n", - "\n", - "You will upload documentation and test results unique to your model based on this template later on. For now, **take a look at the default structure that the template provides with [the `vm.preview_template()` function](https://docs.validmind.ai/validmind/validmind.html#preview_template)** from the ValidMind library and note the empty sections:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "fd332a9d", - "metadata": {}, - "outputs": [], - "source": [ - "vm.preview_template()" - ] - }, - { - "cell_type": "markdown", - "id": "f805ec38", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Initialize the Python environment\n", - "\n", - "Next, let's import the necessary libraries and set up your Python environment for data analysis:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "8e2127cd", - "metadata": {}, - "outputs": [], - "source": [ - "import xgboost as xgb\n", - "\n", - "%matplotlib inline" - ] - }, - { - "cell_type": "markdown", - "id": "1783e13c", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Explore a ValidMind test\n", - "\n", - "Before we run a test, use [the `vm.tests.list_tests()` function](https://docs.validmind.ai/validmind/validmind/tests.html#list_tests) to return information on out-of-the-box tests available in the ValidMind Library.\n", - "\n", - "Let's assume you want to evaluate *classifier performance* for a model. Classifier performance measures how well a classification model correctly predicts outcomes, using metrics like [precision, recall, and F1 score](https://en.wikipedia.org/wiki/Precision_and_recall).\n", - "\n", - "We'll pass in a `filter` to the `list_tests` function to find the test ID for classifier performance:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "a6a6f715", - "metadata": {}, - "outputs": [], - "source": [ - "vm.tests.list_tests(filter=\"ClassifierPerformance\")" - ] - }, - { - "cell_type": "markdown", - "id": "96a56e4b", - "metadata": {}, - "source": [ - "We've identified from the output that the test ID for the classifier performance test is `validmind.model_validation.ClassifierPerformance`.\n", - "\n", - "Use this ID combined with [the `describe_test()` function](https://docs.validmind.ai/validmind/validmind/tests.html#describe_test) to retrieve more information about the test, including its **Required Inputs**:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "f8a46c7d", - "metadata": {}, - "outputs": [], - "source": [ - "test_id = \"validmind.model_validation.sklearn.ClassifierPerformance\"\n", - "vm.tests.describe_test(test_id)" - ] - }, - { - "cell_type": "markdown", - "id": "97053f50", - "metadata": {}, - "source": [ - "Since this test requires a dataset and a model, you can expect it to throw an error when we run it without passing in either as input:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "f853c272", - "metadata": {}, - "outputs": [], - "source": [ - "try:\n", - " vm.tests.run_test(test_id)\n", - "except Exception as e:\n", - " print(e)" - ] - }, - { - "cell_type": "markdown", - "id": "1a3115ed", - "metadata": {}, - "source": [ - "
Learn more about the individual tests available in the ValidMind Library\n", - "

\n", - "Check out our Explore tests notebook for more code examples and usage of key functions.
" - ] - }, - { - "cell_type": "markdown", - "id": "89da851b", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Working with ValidMind datasets" - ] - }, - { - "cell_type": "markdown", - "id": "50bfdb1b", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Import the sample dataset\n", - "\n", - "Since we need a dataset to run tests, let's import the public [Bank Customer Churn Prediction](https://www.kaggle.com/datasets/shantanudhakadd/bank-customer-churn-prediction) dataset from Kaggle so that we have something to work with.\n", - "\n", - "In our below example, note that:\n", - "\n", - "- The target column, `Exited` has a value of `1` when a customer has churned and `0` otherwise.\n", - "- The ValidMind Library provides a wrapper to automatically load the dataset as a [Pandas DataFrame](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.html) object. A Pandas Dataframe is a two-dimensional tabular data structure that makes use of rows and columns." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "3ef2dfbb", - "metadata": {}, - "outputs": [], - "source": [ - "# Import the sample dataset from the library\n", - "\n", - "from validmind.datasets.classification import customer_churn\n", - "\n", - "print(\n", - " f\"Loaded demo dataset with: \\n\\n\\t• Target column: '{customer_churn.target_column}' \\n\\t• Class labels: {customer_churn.class_labels}\"\n", - ")\n", - "\n", - "raw_df = customer_churn.load_data()\n", - "raw_df.head()" - ] - }, - { - "cell_type": "markdown", - "id": "a5a8212f", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Split the dataset\n", - "\n", - "Let's first split our dataset to help assess how well the model generalizes to unseen data.\n", - "\n", - "Use [`preprocess()`](https://docs.validmind.ai/validmind/validmind/datasets/classification/customer_churn.html#preprocess) to split our dataset into three subsets:\n", - "\n", - "1. **train_df** — Used to train the model.\n", - "2. **validation_df** — Used to evaluate the model's performance during training.\n", - "3. **test_df** — Used later on to asses the model's performance on new, unseen data." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "88c87d4a", - "metadata": {}, - "outputs": [], - "source": [ - "train_df, validation_df, test_df = customer_churn.preprocess(raw_df)" - ] - }, - { - "cell_type": "markdown", - "id": "2ae225d7", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Initialize the ValidMind dataset\n", - "\n", - "The next step is to connect your data with a ValidMind `Dataset` object. **This step is always necessary every time you want to connect a dataset to documentation and produce test results through ValidMind,** but you only need to do it once per dataset.\n", - "\n", - "ValidMind dataset objects provide a wrapper to any type of dataset (NumPy, Pandas, Polars, etc.) so that tests can run transparently regardless of the underlying library.\n", - "\n", - "Initialize a ValidMind dataset object using the [`init_dataset` function](https://docs.validmind.ai/validmind/validmind.html#init_dataset) from the ValidMind (`vm`) module. For this example, we'll pass in the following arguments:\n", - "\n", - "- **`dataset`** — The raw dataset that you want to provide as input to tests.\n", - "- **`input_id`** — A unique identifier that allows tracking what inputs are used when running each individual test.\n", - "- **`target_column`** — A required argument if tests require access to true values. This is the name of the target column in the dataset." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "bf0ec747", - "metadata": {}, - "outputs": [], - "source": [ - "vm_train_ds = vm.init_dataset(\n", - " dataset=train_df,\n", - " input_id=\"train_dataset\",\n", - " target_column=customer_churn.target_column,\n", - ")\n", - "\n", - "vm_test_ds = vm.init_dataset(\n", - " dataset=test_df,\n", - " input_id=\"test_dataset\",\n", - " target_column=customer_churn.target_column,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "6d26f65b", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Working with ValidMind models" - ] - }, - { - "cell_type": "markdown", - "id": "6d1677f6", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Train a sample model\n", - "\n", - "To train the model, we need to provide it with:\n", - "\n", - "1. **Inputs** — Features such as customer age, usage, etc.\n", - "2. **Outputs (Expected answers/labels)** — in our case, we would like to know whether the customer churned or not.\n", - "\n", - "Here, we'll use `x_train` and `x_val` to hold the input data (features), and `y_train` and `y_val` to hold the answers (the target we want to predict):" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "39e8c7ea", - "metadata": {}, - "outputs": [], - "source": [ - "x_train = train_df.drop(customer_churn.target_column, axis=1)\n", - "y_train = train_df[customer_churn.target_column]\n", - "x_val = validation_df.drop(customer_churn.target_column, axis=1)\n", - "y_val = validation_df[customer_churn.target_column]" - ] - }, - { - "cell_type": "markdown", - "id": "4ac628eb", - "metadata": {}, - "source": [ - "Next, let's create an *XGBoost classifier model* that will automatically stop training if it doesn't improve after 10 tries. XGBoost is a gradient-boosted tree ensemble that builds trees sequentially, with each tree correcting the errors of the previous ones — typically known for strong predictive performance and built-in regularization to reduce overfitting.\n", - "\n", - "Setting an explicit threshold avoids wasting time and helps prevent further overfitting by stopping training when further improvement isn't happening. We'll also set three evaluation metrics to get a more complete picture of model performance:\n", - "\n", - "1. **error** — Measures how often the model makes incorrect predictions.\n", - "2. **logloss** — Indicates how confident the predictions are.\n", - "3. **auc** — Evaluates how well the model distinguishes between churn and not churn." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "255e3583", - "metadata": {}, - "outputs": [], - "source": [ - "model = xgb.XGBClassifier(early_stopping_rounds=10)\n", - "model.set_params(\n", - " eval_metric=[\"error\", \"logloss\", \"auc\"],\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "f6430312", - "metadata": {}, - "source": [ - "Finally, our actual training step — where the model learns patterns from the data, so it can make predictions later:\n", - "\n", - "- The model is trained on `x_train` and `y_train`, and evaluates its performance using `x_val` and `y_val` to check if it’s learning well.\n", - "- To turn off printed output while training, we'll set `verbose` to `False`." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "e3aa3657", - "metadata": {}, - "outputs": [], - "source": [ - "model.fit(\n", - " x_train,\n", - " y_train,\n", - " eval_set=[(x_val, y_val)],\n", - " verbose=False,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "c303a046", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Initialize the ValidMind model\n", - "\n", - "You'll also need to initialize a ValidMind model object (`vm_model`) that can be passed to other functions for analysis and tests on the data for our model.\n", - "\n", - "You simply initialize this model object with [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model):" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "4b2be11f", - "metadata": {}, - "outputs": [], - "source": [ - "vm_model_xgb = vm.init_model(\n", - " model,\n", - " input_id=\"xgboost\",\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "2fa83857", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Assign predictions\n", - "\n", - "Once the model has been registered, you can assign model predictions to the training and testing datasets.\n", - "\n", - "- The [`assign_predictions()` method](https://docs.validmind.ai/validmind/validmind/vm_models.html#assign_predictions) from the `Dataset` object can link existing predictions to any number of models.\n", - "- This method links the model's class prediction values and probabilities to our `vm_train_ds` and `vm_test_ds` datasets.\n", - "\n", - "If no prediction values are passed, the method will compute predictions automatically:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "229185fd", - "metadata": {}, - "outputs": [], - "source": [ - "vm_train_ds.assign_predictions(model=vm_model_xgb)\n", - "vm_test_ds.assign_predictions(model=vm_model_xgb)" - ] - }, - { - "cell_type": "markdown", - "id": "d0b3312e", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Running ValidMind tests\n", - "\n", - "Now that we know how to initialize ValidMind `dataset` and `model` objects, we're ready to run some tests!\n", - "\n", - "You run individual tests by calling [the `run_test` function](https://docs.validmind.ai/validmind/validmind/tests.html#run_test) provided by the `validmind.tests` module. For the examples below, we'll pass in the following arguments:\n", - "\n", - "- **`test_id`** — The ID of the test to run, as seen in the `ID` column when you run `list_tests`.\n", - "- **`inputs`** — A dictionary of test inputs, such as `dataset`, `model`, `datasets`, or `models`. These are ValidMind objects initialized with [`vm.init_dataset()`](https://docs.validmind.ai/validmind/validmind.html#init_dataset) or [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model)." - ] - }, - { - "cell_type": "markdown", - "id": "96c89f32", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Run classifier performance test with one model\n", - "\n", - "Run `validmind.data_validation.ClassifierPerformance` test with the testing dataset (`vm_test_ds`) and model (`vm_model_xgb`) as inputs:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "85189af9", - "metadata": {}, - "outputs": [], - "source": [ - "result = vm.tests.run_test(\n", - " \"validmind.model_validation.sklearn.ClassifierPerformance\",\n", - " inputs={\n", - " \"dataset\": vm_test_ds,\n", - " \"model\": vm_model_xgb,\n", - " },\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "676dff89", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Run comparison tests\n", - "\n", - "To evaluate which models might be a better fit for a use case based on their performance on selected criteria, we can run the same test with multiple models. We'll train three additional models and run the classifier performance test with for all four models using a single `run_test()` call.\n", - "\n", - "
ValidMind helps streamline your documentation and testing.\n", - "

\n", - "You could call run_test() multiple times passing in different inputs, but you can also pass an input_grid object — a dictionary of test input keys and values that allow you to run a single test for a combination of models and datasets.\n", - "

\n", - "With input_grid, run comparison tests for multiple datasets, or even multiple datasets and models simultaneously — input_grid can be used with run_test() for all possible combinations of inputs, generating a cohesive and comprehensive single output.\n", - "
" - ] - }, - { - "cell_type": "markdown", - "id": "3d9912dc", - "metadata": {}, - "source": [ - "*Random forest classifier* models use an ensemble method that builds multiple decision trees and averages their predictions. Random forest is robust to overfitting and handles non-linear relations well, but is typically less interpretable than simpler models:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "1976b7e8", - "metadata": {}, - "outputs": [], - "source": [ - "from sklearn.ensemble import RandomForestClassifier\n", - "\n", - "# Train the random forest classifer model\n", - "model_rf = RandomForestClassifier()\n", - "model_rf.fit(x_train, y_train)\n", - "\n", - "# Initialize the ValidMind model object for the random forest classifer model\n", - "vm_model_rf = vm.init_model(\n", - " model_rf,\n", - " input_id=\"random_forest\",\n", - ")\n", - "\n", - "# Assign predictions to the test dataset for the random forest classifer model\n", - "vm_test_ds.assign_predictions(model=vm_model_rf)" - ] - }, - { - "cell_type": "markdown", - "id": "a259927c", - "metadata": {}, - "source": [ - "*Logistic regression* models are linear models that estimate class probabilities via a logistic (sigmoid) function. Logistic regression is highly interpretable with fast training, establishing a strong baseline — however, they struggle when relationships are non-linear as real-world relationships often are:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "90bbf148", - "metadata": {}, - "outputs": [], - "source": [ - "from sklearn.linear_model import LogisticRegression\n", - "from sklearn.preprocessing import StandardScaler\n", - "from sklearn.pipeline import Pipeline\n", - "\n", - "# Scaling features ensures the lbfgs solver converges reliably\n", - "model_lr = Pipeline([\n", - " (\"scaler\", StandardScaler()),\n", - " (\"lr\", LogisticRegression()),\n", - "])\n", - "model_lr.fit(x_train, y_train)\n", - "\n", - "# Initialize the ValidMind model object for the logistic regression model\n", - "vm_model_lr = vm.init_model(\n", - " model_lr,\n", - " input_id=\"logistic_regression\",\n", - ")\n", - "\n", - "# Assign predictions to the test dataset for the logistic regression model\n", - "vm_test_ds.assign_predictions(model=vm_model_lr)" - ] - }, - { - "cell_type": "markdown", - "id": "9a666b41", - "metadata": {}, - "source": [ - "*Decision tree classifier* models are a single tree with data split on feature thresholds. Useful as an explanability benchmark, decision trees are easy to visualize and interpret — but are prone to overfitting without pruning or ensemble techniques:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "bfa1e17d", - "metadata": {}, - "outputs": [], - "source": [ - "from sklearn.tree import DecisionTreeClassifier\n", - "\n", - "# Train the decision tree classifer model\n", - "model_dt = DecisionTreeClassifier()\n", - "model_dt.fit(x_train, y_train)\n", - "\n", - "# Initialize the ValidMind model object for the decision tree classifier model\n", - "vm_model_dt = vm.init_model(\n", - " model_dt,\n", - " input_id=\"decision_tree\",\n", - ")\n", - "\n", - "# Assign predictions to the test dataset for the decision tree classifiermodel\n", - "vm_test_ds.assign_predictions(model=vm_model_dt)" - ] - }, - { - "cell_type": "markdown", - "id": "2c8f3268", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Run classifier performance test with multiple models\n", - "\n", - "Now, we'll use the `input_grid` to run the [`ClassifierPerformance` test](https://docs.validmind.ai/tests/model_validation/sklearn/ClassifierPerformance.html) on all four models using the testing dataset (`vm_test_ds`).\n", - "\n", - "When running individual tests, you can use a custom `result_id` to tag the individual result with a unique identifier by appending this `result_id` to the `test_id` with a `:` separator. We'll append an identifier to signify that this test was run on `all_models` to differentiate this test run from other runs:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "2e48ce1e", - "metadata": {}, - "outputs": [], - "source": [ - "perf_comparison_result = vm.tests.run_test(\n", - " \"validmind.model_validation.sklearn.ClassifierPerformance:all_models\",\n", - " input_grid={\n", - " \"dataset\": [vm_test_ds],\n", - " \"model\": [vm_model_xgb, vm_model_rf, vm_model_lr, vm_model_dt],\n", - " },\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "81cbf144", - "metadata": {}, - "source": [ - "Our output indicates that the XGBoost and random forest classification models provide the strongest overall classification performance, so we'll continue our testing with those two models as input only." - ] - }, - { - "cell_type": "markdown", - "id": "3d3fb6ec", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Run classifier performance test with multiple parameter values\n", - "\n", - "Next, let's run the classifier performance test with the `param_grid` object, which runs the same test multiple times with different parameter values. We'll append an identifier to signify that this test was run with our `parameter_grid` configuration:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "d0ad94c9", - "metadata": {}, - "outputs": [], - "source": [ - "parameter_comparison_result = vm.tests.run_test(\n", - " \"validmind.model_validation.sklearn.ClassifierPerformance:parameter_grid\",\n", - " input_grid={\n", - " \"dataset\": [vm_test_ds],\n", - " \"model\": [vm_model_xgb,vm_model_rf]\n", - " },\n", - " param_grid={\n", - " \"average\": [\"macro\", \"micro\"]\n", - " },\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "508c7546", - "metadata": {}, - "source": [ - "\n", - "\n", - "#### Run comparison test with multiple datasets\n", - "\n", - "Let's also run the [ROCCurve test](https://docs.validmind.ai/tests/model_validation/sklearn/ROCCurve.html) using `input_grid` to iterate through multiple datasets, which plots the ROC curves for the training (`vm_train_ds`) and test (`vm_test_ds`) datasets side by side — a common scenario when you want to compare the performance of a model on the training and test datasets and visually assess how much performance is lost in the test dataset.\n", - "\n", - "We'll also need to assign predictions to the training dataset for the random forest classifier model, since we didn't do that in our earlier setup:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "96c3b426", - "metadata": {}, - "outputs": [], - "source": [ - "vm_train_ds.assign_predictions(model=vm_model_rf)" - ] - }, - { - "cell_type": "markdown", - "id": "2be82bae", - "metadata": {}, - "source": [ - "We'll append an identifier to signify that this test was run with our `train_vs_test` dataset comparison configuration:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "4056aa1e", - "metadata": {}, - "outputs": [], - "source": [ - "roc_curve_result = vm.tests.run_test(\n", - " \"validmind.model_validation.sklearn.ROCCurve:train_vs_test\",\n", - " input_grid={\n", - " \"dataset\": [vm_train_ds, vm_test_ds],\n", - " \"model\": [vm_model_xgb,vm_model_rf],\n", - " },\n", - ")" - ] - }, - { - "cell_type": "markdown", - "id": "a05570d5", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Work with test results\n", - "\n", - "Every test result returned by the `run_test()` function has a [`.log()` method](https://docs.validmind.ai/validmind/validmind/vm_models.html#TestResult.log) that can be used to send the test results to the ValidMind Platform. When logging individual test results to the platform, you'll need to manually add those results to the desired section of the model documentation.\n", - "\n", - "You can do this through the ValidMind Platform interface after logging your test results ([Learn more ...](https://docs.validmind.ai/developer/model-documentation/work-with-test-results.html)), or directly via the ValidMind Library when calling `.log()` by providing an optional `section_id`. The `section_id` should be a string that matches the title of a section in the documentation template in `snake_case`.\n", - "\n", - "Let's log the results of the classifier performance test (`perf_comparison_result`) and the ROCCurve (`roc_curve_result`) test in the `model_evaluation` section of the documentation — present in the template we previewed in the beginning of this notebook:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "e119bf1e", - "metadata": {}, - "outputs": [], - "source": [ - "perf_comparison_result.log(section_id=\"model_evaluation\")\n", - "roc_curve_result.log(section_id=\"model_evaluation\")" - ] - }, - { - "cell_type": "markdown", - "id": "ab5205ee", - "metadata": {}, - "source": [ - "Finally, let's head to the model we connected to at the beginning of this notebook and view our inserted test results in the updated documentation ([Need more help?](https://docs.validmind.ai/guide/model-documentation/working-with-model-documentation.html)):\n", - "\n", - "1. From the **Inventory** in the ValidMind Platform, go to the model you connected to earlier.\n", - "\n", - "2. In the left sidebar that appears for your model, click **Development** under Documents.\n", - "\n", - "3. Expand the **3.2. Model Evaluation** section.\n", - "\n", - "4. Confirm that `perf_comparison_result` and `roc_curve_result` display in this section as expected." - ] - }, - { - "cell_type": "markdown", - "id": "eb196aac", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Next steps\n", - "\n", - "Now that you know how to run comparison tests with the ValidMind Library, you’re ready to take the next step. Extend the functionality of `run_test()` with your own custom test functions that can be incorporated into documentation templates just like any default out-of-the-box ValidMind test.\n", - "\n", - "
Learn how to implement custom tests with the ValidMind Library.\n", - "

\n", - "Check out our Implement comparison tests notebook for code examples and usage of key functions.
" - ] - }, - { - "cell_type": "markdown", - "id": "083c1d8d", - "metadata": {}, - "source": [ - "\n", - "\n", - "### Discover more learning resources\n", - "\n", - "We offer many interactive notebooks to help you automate testing, documenting, validating, and more:\n", - "\n", - "- [Run tests & test suites](https://docs.validmind.ai/developer/how-to/testing-overview.html)\n", - "- [Use ValidMind Library features](https://docs.validmind.ai/developer/how-to/feature-overview.html)\n", - "- [Code samples by use case](https://docs.validmind.ai/guide/samples-jupyter-notebooks.html)\n", - "\n", - "Or, visit our [documentation](https://docs.validmind.ai/) to learn more about ValidMind." - ] - }, - { - "cell_type": "markdown", - "id": "efba0f57", - "metadata": {}, - "source": [ - "\n", - "\n", - "## Upgrade ValidMind\n", - "\n", - "
After installing ValidMind, you’ll want to periodically make sure you are on the latest version to access any new features and other enhancements.
\n", - "\n", - "Retrieve the information for the currently installed version of ValidMind:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "0d35972c", - "metadata": { - "vscode": { - "languageId": "plaintext" - } - }, - "outputs": [], - "source": [ - "%pip show validmind" - ] - }, - { - "cell_type": "markdown", - "id": "abcd07ef", - "metadata": {}, - "source": [ - "If the version returned is lower than the version indicated in our [production open-source code](https://github.com/validmind/validmind-library/blob/prod/validmind/__version__.py), restart your notebook and run:\n", - "\n", - "```bash\n", - "%pip install --upgrade validmind\n", - "```" - ] - }, - { - "cell_type": "markdown", - "id": "5fe70b90", - "metadata": {}, - "source": [ - "You may need to restart your kernel after running the upgrade package for changes to be applied." - ] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [ - "\n", - "\n", - "\n", - "\n", - "***\n", - "\n", - "Copyright © 2023-2026 ValidMind Inc. All rights reserved.
\n", - "Refer to [LICENSE](https://github.com/validmind/validmind-library/blob/main/LICENSE) for details.
\n", - "SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial" - ] - } - ], - "metadata": { - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - }, - "language_info": { - "name": "python", - "version": "3.10" + "cells": [ + { + "cell_type": "markdown", + "id": "1d29276f", + "metadata": {}, + "source": [ + "# Run comparison tests\n", + "\n", + "Learn how to use the ValidMind Library to run comparison tests that take any datasets or models as inputs. Identify comparison tests to run, initialize ValidMind dataset and model objects in preparation for passing them to tests, and then run tests — generating outputs automatically logged to your model's documentation in the ValidMind Platform.\n", + "\n", + "
We recommend that you first complete our introductory notebook on running tests.\n", + "

\n", + "Run dataset-based tests
" + ] + }, + { + "cell_type": "markdown", + "id": "61065444", + "metadata": {}, + "source": [ + "::: {.content-hidden when-format=\"html\"}\n", + "## Contents \n", + "- [About ValidMind](#toc1__) \n", + " - [Before you begin](#toc1_1__) \n", + " - [New to ValidMind?](#toc1_2__) \n", + " - [Key concepts](#toc1_3__) \n", + "- [Setting up](#toc2__) \n", + " - [Install the ValidMind Library](#toc2_1__) \n", + " - [Initialize the ValidMind Library](#toc2_2__) \n", + " - [Register sample model](#toc2_2_1__) \n", + " - [Apply documentation template](#toc2_2_2__) \n", + " - [Get your code snippet](#toc2_2_3__) \n", + " - [Preview the documentation template](#toc2_3__) \n", + " - [Initialize the Python environment](#toc2_4__) \n", + "- [Explore a ValidMind test](#toc3__) \n", + "- [Working with ValidMind datasets](#toc4__) \n", + " - [Import the sample dataset](#toc4_1__) \n", + " - [Split the dataset](#toc4_2__) \n", + " - [Initialize the ValidMind dataset](#toc4_3__) \n", + "- [Working with ValidMind models](#toc5__) \n", + " - [Train a sample model](#toc5_1__) \n", + " - [Initialize the ValidMind model](#toc5_2__) \n", + " - [Assign predictions](#toc5_3__) \n", + "- [Running ValidMind tests](#toc6__) \n", + " - [Run classifier performance test with one model](#toc6_1__) \n", + " - [Run comparison tests](#toc6_2__) \n", + " - [Run classifier performance test with multiple models](#toc6_2_1__) \n", + " - [Run classifier performance test with multiple parameter values](#toc6_2_2__) \n", + " - [Run comparison test with multiple datasets](#toc6_2_3__) \n", + "- [Work with test results](#toc7__) \n", + "- [Next steps](#toc8__) \n", + " - [Discover more learning resources](#toc8_1__) \n", + "- [Upgrade ValidMind](#toc9__) \n", + "\n", + ":::\n", + "\n", + "" + ] + }, + { + "cell_type": "markdown", + "id": "67a4d9dc", + "metadata": {}, + "source": [ + "\n", + "\n", + "## About ValidMind\n", + "\n", + "ValidMind is a suite of tools for managing model risk, including risk associated with AI and statistical models. \n", + "\n", + "You use the ValidMind Library to automate documentation and validation tests, and then use the ValidMind Platform to collaborate on model documentation. Together, these products simplify model risk management, facilitate compliance with regulations and institutional standards, and enhance collaboration between yourself and model validators." + ] + }, + { + "cell_type": "markdown", + "id": "eeb30df8", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Before you begin\n", + "\n", + "This notebook assumes you have basic familiarity with Python, including an understanding of how functions work. If you are new to Python, you can still run the notebook but we recommend further familiarizing yourself with the language. \n", + "\n", + "If you encounter errors due to missing modules in your Python environment, install the modules with `pip install`, and then re-run the notebook. For more help, refer to [Installing Python Modules](https://docs.python.org/3/installing/index.html)." + ] + }, + { + "cell_type": "markdown", + "id": "293c3f98", + "metadata": {}, + "source": [ + "\n", + "\n", + "### New to ValidMind?\n", + "\n", + "If you haven't already seen our documentation on the [ValidMind Library](https://docs.validmind.ai/developer/validmind-library.html), we recommend you begin by exploring the available resources in this section. There, you can learn more about documenting models and running tests, as well as find code samples and our Python Library API reference.\n", + "\n", + "
For access to all features available in this notebook, you'll need access to a ValidMind account.\n", + "

\n", + "Register with ValidMind
" + ] + }, + { + "cell_type": "markdown", + "id": "4fc836d0", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Key concepts\n", + "\n", + "**Model documentation**: A structured and detailed record pertaining to a model, encompassing key components such as its underlying assumptions, methodologies, data sources, inputs, performance metrics, evaluations, limitations, and intended uses. It serves to ensure transparency, adherence to regulatory requirements, and a clear understanding of potential risks associated with the model’s application.\n", + "\n", + "**Documentation template**: Functions as a test suite and lays out the structure of model documentation, segmented into various sections and sub-sections. Documentation templates define the structure of your model documentation, specifying the tests that should be run, and how the results should be displayed.\n", + "\n", + "**Tests**: A function contained in the ValidMind Library, designed to run a specific quantitative test on the dataset or model. Tests are the building blocks of ValidMind, used to evaluate and document models and datasets, and can be run individually or as part of a suite defined by your model documentation template.\n", + "\n", + "**Metrics**: A subset of tests that do not have thresholds. In the context of this notebook, metrics and tests can be thought of as interchangeable concepts.\n", + "\n", + "**Custom metrics**: Custom metrics are functions that you define to evaluate your model or dataset. These functions can be registered with the ValidMind Library to be used in the ValidMind Platform.\n", + "\n", + "**Inputs**: Objects to be evaluated and documented in the ValidMind Library. They can be any of the following:\n", + "\n", + " - **model**: A single model that has been initialized in ValidMind with [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model).\n", + " - **dataset**: Single dataset that has been initialized in ValidMind with [`vm.init_dataset()`](https://docs.validmind.ai/validmind/validmind.html#init_dataset).\n", + " - **models**: A list of ValidMind models - usually this is used when you want to compare multiple models in your custom metric.\n", + " - **datasets**: A list of ValidMind datasets - usually this is used when you want to compare multiple datasets in your custom metric. (Learn more: [Run tests with multiple datasets](https://docs.validmind.ai/notebooks/how_to/tests/run_tests/configure_tests/run_tests_that_require_multiple_datasets.html))\n", + "\n", + "**Parameters**: Additional arguments that can be passed when running a ValidMind test, used to pass additional information to a metric, customize its behavior, or provide additional context.\n", + "\n", + "**Outputs**: Custom metrics can return elements like tables or plots. Tables may be a list of dictionaries (each representing a row) or a pandas DataFrame. Plots may be matplotlib or plotly figures.\n", + "\n", + "**Test suites**: Collections of tests designed to run together to automate and generate model documentation end-to-end for specific use-cases.\n", + "\n", + "Example: the [`classifier_full_suite`](https://docs.validmind.ai/validmind/validmind/test_suites/classifier.html#ClassifierFullSuite) test suite runs tests from the [`tabular_dataset`](https://docs.validmind.ai/validmind/validmind/test_suites/tabular_datasets.html) and [`classifier`](https://docs.validmind.ai/validmind/validmind/test_suites/classifier.html) test suites to fully document the data and model sections for binary classification model use-cases." + ] + }, + { + "cell_type": "markdown", + "id": "8d52b6e0", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Setting up" + ] + }, + { + "cell_type": "markdown", + "id": "e0d2daaf", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Install the ValidMind Library\n", + "\n", + "
Recommended Python versions\n", + "

\n", + "Python 3.8 <= x <= 3.14
\n", + "\n", + "To install the library:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc97888f", + "metadata": {}, + "outputs": [], + "source": [ + "%pip install -q validmind" + ] + }, + { + "cell_type": "markdown", + "id": "1ff56571", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Initialize the ValidMind Library" + ] + }, + { + "cell_type": "markdown", + "id": "c4d9f164", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Register sample model\n", + "\n", + "Let's first register a sample model for use with this notebook.\n", + "\n", + "1. In a browser, [log in to ValidMind](https://docs.validmind.ai/guide/configuration/log-in-to-validmind.html).\n", + "\n", + "2. In the left sidebar, navigate to **Inventory** and click **+ Register Model**.\n", + "\n", + "3. Enter the model details and click **Next >** to continue to assignment of model stakeholders. ([Need more help?](https://docs.validmind.ai/guide/model-inventory/register-models-in-inventory.html))\n", + "\n", + "4. Select your own name under the **MODEL OWNER** drop-down.\n", + "\n", + "5. Click **Register Model** to add the model to your inventory." + ] + }, + { + "cell_type": "markdown", + "id": "852392e5", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Apply documentation template\n", + "\n", + "Once you've registered your model, let's select a documentation template. A template predefines sections for your model documentation and provides a general outline to follow, making the documentation process much easier.\n", + "\n", + "1. In the left sidebar that appears for your model, click **Documents** and select **Development**.\n", + "\n", + "2. Under **TEMPLATE**, select `Binary classification`.\n", + "\n", + "3. Click **Use Template** to apply the template." + ] + }, + { + "cell_type": "markdown", + "id": "6490e991", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Get your code snippet\n", + "\n", + "Initialize the ValidMind Library with the *code snippet* unique to each model per document, ensuring your test results are uploaded to the correct model and automatically populated in the right document in the ValidMind Platform when you run this notebook.\n", + "\n", + "1. On the left sidebar that appears for your model, select **Getting Started** and select `Development` from the **DOCUMENT** drop-down menu.\n", + "2. Click **Copy snippet to clipboard**.\n", + "3. Next, [load your model identifier credentials from an `.env` file](https://docs.validmind.ai/developer/model-documentation/store-credentials-in-env-file.html) or replace the placeholder with your own code snippet::" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c51ae01c", + "metadata": {}, + "outputs": [], + "source": [ + "# Load your model identifier credentials from an `.env` file\n", + "\n", + "%load_ext dotenv\n", + "%dotenv .env\n", + "\n", + "# Or replace with your code snippet\n", + "\n", + "import validmind as vm\n", + "\n", + "vm.init(\n", + " # api_host=\"...\",\n", + " # api_key=\"...\",\n", + " # api_secret=\"...\",\n", + " # model=\"...\",\n", + " document=\"documentation\",\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "99e9d14f", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Preview the documentation template\n", + "\n", + "Let's verify that you have connected the ValidMind Library to the ValidMind Platform and that the appropriate *template* is selected for your model.\n", + "\n", + "You will upload documentation and test results unique to your model based on this template later on. For now, **take a look at the default structure that the template provides with [the `vm.preview_template()` function](https://docs.validmind.ai/validmind/validmind.html#preview_template)** from the ValidMind library and note the empty sections:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fd332a9d", + "metadata": {}, + "outputs": [], + "source": [ + "vm.preview_template()" + ] + }, + { + "cell_type": "markdown", + "id": "f805ec38", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Initialize the Python environment\n", + "\n", + "Next, let's import the necessary libraries and set up your Python environment for data analysis:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8e2127cd", + "metadata": {}, + "outputs": [], + "source": [ + "import xgboost as xgb\n", + "\n", + "%matplotlib inline" + ] + }, + { + "cell_type": "markdown", + "id": "1783e13c", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Explore a ValidMind test\n", + "\n", + "Before we run a test, use [the `vm.tests.list_tests()` function](https://docs.validmind.ai/validmind/validmind/tests.html#list_tests) to return information on out-of-the-box tests available in the ValidMind Library.\n", + "\n", + "Let's assume you want to evaluate *classifier performance* for a model. Classifier performance measures how well a classification model correctly predicts outcomes, using metrics like [precision, recall, and F1 score](https://en.wikipedia.org/wiki/Precision_and_recall).\n", + "\n", + "We'll pass in a `filter` to the `list_tests` function to find the test ID for classifier performance:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a6a6f715", + "metadata": {}, + "outputs": [], + "source": [ + "vm.tests.list_tests(filter=\"ClassifierPerformance\")" + ] + }, + { + "cell_type": "markdown", + "id": "96a56e4b", + "metadata": {}, + "source": [ + "We've identified from the output that the test ID for the classifier performance test is `validmind.model_validation.ClassifierPerformance`.\n", + "\n", + "Use this ID combined with [the `describe_test()` function](https://docs.validmind.ai/validmind/validmind/tests.html#describe_test) to retrieve more information about the test, including its **Required Inputs**:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f8a46c7d", + "metadata": {}, + "outputs": [], + "source": [ + "test_id = \"validmind.model_validation.sklearn.ClassifierPerformance\"\n", + "vm.tests.describe_test(test_id)" + ] + }, + { + "cell_type": "markdown", + "id": "97053f50", + "metadata": {}, + "source": [ + "Since this test requires a dataset and a model, you can expect it to throw an error when we run it without passing in either as input:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f853c272", + "metadata": {}, + "outputs": [], + "source": [ + "try:\n", + " vm.tests.run_test(test_id)\n", + "except Exception as e:\n", + " print(e)" + ] + }, + { + "cell_type": "markdown", + "id": "1a3115ed", + "metadata": {}, + "source": [ + "
Learn more about the individual tests available in the ValidMind Library\n", + "

\n", + "Check out our Explore tests notebook for more code examples and usage of key functions.
" + ] + }, + { + "cell_type": "markdown", + "id": "89da851b", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Working with ValidMind datasets" + ] + }, + { + "cell_type": "markdown", + "id": "50bfdb1b", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Import the sample dataset\n", + "\n", + "Since we need a dataset to run tests, let's import the public [Bank Customer Churn Prediction](https://www.kaggle.com/datasets/shantanudhakadd/bank-customer-churn-prediction) dataset from Kaggle so that we have something to work with.\n", + "\n", + "In our below example, note that:\n", + "\n", + "- The target column, `Exited` has a value of `1` when a customer has churned and `0` otherwise.\n", + "- The ValidMind Library provides a wrapper to automatically load the dataset as a [Pandas DataFrame](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.html) object. A Pandas Dataframe is a two-dimensional tabular data structure that makes use of rows and columns." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3ef2dfbb", + "metadata": {}, + "outputs": [], + "source": [ + "# Import the sample dataset from the library\n", + "\n", + "from validmind.datasets.classification import customer_churn\n", + "\n", + "print(\n", + " f\"Loaded demo dataset with: \\n\\n\\t• Target column: '{customer_churn.target_column}' \\n\\t• Class labels: {customer_churn.class_labels}\"\n", + ")\n", + "\n", + "raw_df = customer_churn.load_data()\n", + "raw_df.head()" + ] + }, + { + "cell_type": "markdown", + "id": "a5a8212f", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Split the dataset\n", + "\n", + "Let's first split our dataset to help assess how well the model generalizes to unseen data.\n", + "\n", + "Use [`preprocess()`](https://docs.validmind.ai/validmind/validmind/datasets/classification/customer_churn.html#preprocess) to split our dataset into three subsets:\n", + "\n", + "1. **train_df** — Used to train the model.\n", + "2. **validation_df** — Used to evaluate the model's performance during training.\n", + "3. **test_df** — Used later on to asses the model's performance on new, unseen data." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "88c87d4a", + "metadata": {}, + "outputs": [], + "source": [ + "train_df, validation_df, test_df = customer_churn.preprocess(raw_df)" + ] + }, + { + "cell_type": "markdown", + "id": "2ae225d7", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Initialize the ValidMind dataset\n", + "\n", + "The next step is to connect your data with a ValidMind `Dataset` object. **This step is always necessary every time you want to connect a dataset to documentation and produce test results through ValidMind,** but you only need to do it once per dataset.\n", + "\n", + "ValidMind dataset objects provide a wrapper to any type of dataset (NumPy, Pandas, Polars, etc.) so that tests can run transparently regardless of the underlying library.\n", + "\n", + "Initialize a ValidMind dataset object using the [`init_dataset` function](https://docs.validmind.ai/validmind/validmind.html#init_dataset) from the ValidMind (`vm`) module. For this example, we'll pass in the following arguments:\n", + "\n", + "- **`dataset`** — The raw dataset that you want to provide as input to tests.\n", + "- **`input_id`** — A unique identifier that allows tracking what inputs are used when running each individual test.\n", + "- **`target_column`** — A required argument if tests require access to true values. This is the name of the target column in the dataset." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bf0ec747", + "metadata": {}, + "outputs": [], + "source": [ + "vm_train_ds = vm.init_dataset(\n", + " dataset=train_df,\n", + " input_id=\"train_dataset\",\n", + " target_column=customer_churn.target_column,\n", + ")\n", + "\n", + "vm_test_ds = vm.init_dataset(\n", + " dataset=test_df,\n", + " input_id=\"test_dataset\",\n", + " target_column=customer_churn.target_column,\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "6d26f65b", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Working with ValidMind models" + ] + }, + { + "cell_type": "markdown", + "id": "6d1677f6", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Train a sample model\n", + "\n", + "To train the model, we need to provide it with:\n", + "\n", + "1. **Inputs** — Features such as customer age, usage, etc.\n", + "2. **Outputs (Expected answers/labels)** — in our case, we would like to know whether the customer churned or not.\n", + "\n", + "Here, we'll use `x_train` and `x_val` to hold the input data (features), and `y_train` and `y_val` to hold the answers (the target we want to predict):" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "39e8c7ea", + "metadata": {}, + "outputs": [], + "source": [ + "x_train = train_df.drop(customer_churn.target_column, axis=1)\n", + "y_train = train_df[customer_churn.target_column]\n", + "x_val = validation_df.drop(customer_churn.target_column, axis=1)\n", + "y_val = validation_df[customer_churn.target_column]" + ] + }, + { + "cell_type": "markdown", + "id": "4ac628eb", + "metadata": {}, + "source": [ + "Next, let's create an *XGBoost classifier model* that will automatically stop training if it doesn't improve after 10 tries. XGBoost is a gradient-boosted tree ensemble that builds trees sequentially, with each tree correcting the errors of the previous ones — typically known for strong predictive performance and built-in regularization to reduce overfitting.\n", + "\n", + "Setting an explicit threshold avoids wasting time and helps prevent further overfitting by stopping training when further improvement isn't happening. We'll also set three evaluation metrics to get a more complete picture of model performance:\n", + "\n", + "1. **error** — Measures how often the model makes incorrect predictions.\n", + "2. **logloss** — Indicates how confident the predictions are.\n", + "3. **auc** — Evaluates how well the model distinguishes between churn and not churn." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "255e3583", + "metadata": {}, + "outputs": [], + "source": [ + "model = xgb.XGBClassifier(early_stopping_rounds=10)\n", + "model.set_params(\n", + " eval_metric=[\"error\", \"logloss\", \"auc\"],\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "f6430312", + "metadata": {}, + "source": [ + "Finally, our actual training step — where the model learns patterns from the data, so it can make predictions later:\n", + "\n", + "- The model is trained on `x_train` and `y_train`, and evaluates its performance using `x_val` and `y_val` to check if it’s learning well.\n", + "- To turn off printed output while training, we'll set `verbose` to `False`." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e3aa3657", + "metadata": {}, + "outputs": [], + "source": [ + "model.fit(\n", + " x_train,\n", + " y_train,\n", + " eval_set=[(x_val, y_val)],\n", + " verbose=False,\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "c303a046", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Initialize the ValidMind model\n", + "\n", + "You'll also need to initialize a ValidMind model object (`vm_model`) that can be passed to other functions for analysis and tests on the data for our model.\n", + "\n", + "You simply initialize this model object with [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model):" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4b2be11f", + "metadata": {}, + "outputs": [], + "source": [ + "vm_model_xgb = vm.init_model(\n", + " model,\n", + " input_id=\"xgboost\",\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "2fa83857", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Assign predictions\n", + "\n", + "Once the model has been registered, you can assign model predictions to the training and testing datasets.\n", + "\n", + "- The [`assign_predictions()` method](https://docs.validmind.ai/validmind/validmind/vm_models.html#assign_predictions) from the `Dataset` object can link existing predictions to any number of models.\n", + "- This method links the model's class prediction values and probabilities to our `vm_train_ds` and `vm_test_ds` datasets.\n", + "\n", + "If no prediction values are passed, the method will compute predictions automatically:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "229185fd", + "metadata": {}, + "outputs": [], + "source": [ + "vm_train_ds.assign_predictions(model=vm_model_xgb)\n", + "vm_test_ds.assign_predictions(model=vm_model_xgb)" + ] + }, + { + "cell_type": "markdown", + "id": "d0b3312e", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Running ValidMind tests\n", + "\n", + "Now that we know how to initialize ValidMind `dataset` and `model` objects, we're ready to run some tests!\n", + "\n", + "You run individual tests by calling [the `run_test` function](https://docs.validmind.ai/validmind/validmind/tests.html#run_test) provided by the `validmind.tests` module. For the examples below, we'll pass in the following arguments:\n", + "\n", + "- **`test_id`** — The ID of the test to run, as seen in the `ID` column when you run `list_tests`.\n", + "- **`inputs`** — A dictionary of test inputs, such as `dataset`, `model`, `datasets`, or `models`. These are ValidMind objects initialized with [`vm.init_dataset()`](https://docs.validmind.ai/validmind/validmind.html#init_dataset) or [`vm.init_model()`](https://docs.validmind.ai/validmind/validmind.html#init_model)." + ] + }, + { + "cell_type": "markdown", + "id": "96c89f32", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Run classifier performance test with one model\n", + "\n", + "Run `validmind.data_validation.ClassifierPerformance` test with the testing dataset (`vm_test_ds`) and model (`vm_model_xgb`) as inputs:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85189af9", + "metadata": {}, + "outputs": [], + "source": [ + "result = vm.tests.run_test(\n", + " \"validmind.model_validation.sklearn.ClassifierPerformance\",\n", + " inputs={\n", + " \"dataset\": vm_test_ds,\n", + " \"model\": vm_model_xgb,\n", + " },\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "676dff89", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Run comparison tests\n", + "\n", + "To evaluate which models might be a better fit for a use case based on their performance on selected criteria, we can run the same test with multiple models. We'll train three additional models and run the classifier performance test with for all four models using a single `run_test()` call.\n", + "\n", + "
ValidMind helps streamline your documentation and testing.\n", + "

\n", + "You could call run_test() multiple times passing in different inputs, but you can also pass an input_grid object — a dictionary of test input keys and values that allow you to run a single test for a combination of models and datasets.\n", + "

\n", + "With input_grid, run comparison tests for multiple datasets, or even multiple datasets and models simultaneously — input_grid can be used with run_test() for all possible combinations of inputs, generating a cohesive and comprehensive single output.\n", + "
" + ] + }, + { + "cell_type": "markdown", + "id": "3d9912dc", + "metadata": {}, + "source": [ + "*Random forest classifier* models use an ensemble method that builds multiple decision trees and averages their predictions. Random forest is robust to overfitting and handles non-linear relations well, but is typically less interpretable than simpler models:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1976b7e8", + "metadata": {}, + "outputs": [], + "source": [ + "from sklearn.ensemble import RandomForestClassifier\n", + "\n", + "# Train the random forest classifer model\n", + "model_rf = RandomForestClassifier()\n", + "model_rf.fit(x_train, y_train)\n", + "\n", + "# Initialize the ValidMind model object for the random forest classifer model\n", + "vm_model_rf = vm.init_model(\n", + " model_rf,\n", + " input_id=\"random_forest\",\n", + ")\n", + "\n", + "# Assign predictions to the test dataset for the random forest classifer model\n", + "vm_test_ds.assign_predictions(model=vm_model_rf)" + ] + }, + { + "cell_type": "markdown", + "id": "a259927c", + "metadata": {}, + "source": [ + "*Logistic regression* models are linear models that estimate class probabilities via a logistic (sigmoid) function. Logistic regression is highly interpretable with fast training, establishing a strong baseline — however, they struggle when relationships are non-linear as real-world relationships often are:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "90bbf148", + "metadata": {}, + "outputs": [], + "source": [ + "from sklearn.linear_model import LogisticRegression\n", + "from sklearn.preprocessing import StandardScaler\n", + "from sklearn.pipeline import Pipeline\n", + "\n", + "# Scaling features ensures the lbfgs solver converges reliably\n", + "model_lr = Pipeline([\n", + " (\"scaler\", StandardScaler()),\n", + " (\"lr\", LogisticRegression()),\n", + "])\n", + "model_lr.fit(x_train, y_train)\n", + "\n", + "# Initialize the ValidMind model object for the logistic regression model\n", + "vm_model_lr = vm.init_model(\n", + " model_lr,\n", + " input_id=\"logistic_regression\",\n", + ")\n", + "\n", + "# Assign predictions to the test dataset for the logistic regression model\n", + "vm_test_ds.assign_predictions(model=vm_model_lr)" + ] + }, + { + "cell_type": "markdown", + "id": "9a666b41", + "metadata": {}, + "source": [ + "*Decision tree classifier* models are a single tree with data split on feature thresholds. Useful as an explanability benchmark, decision trees are easy to visualize and interpret — but are prone to overfitting without pruning or ensemble techniques:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bfa1e17d", + "metadata": {}, + "outputs": [], + "source": [ + "from sklearn.tree import DecisionTreeClassifier\n", + "\n", + "# Train the decision tree classifer model\n", + "model_dt = DecisionTreeClassifier()\n", + "model_dt.fit(x_train, y_train)\n", + "\n", + "# Initialize the ValidMind model object for the decision tree classifier model\n", + "vm_model_dt = vm.init_model(\n", + " model_dt,\n", + " input_id=\"decision_tree\",\n", + ")\n", + "\n", + "# Assign predictions to the test dataset for the decision tree classifiermodel\n", + "vm_test_ds.assign_predictions(model=vm_model_dt)" + ] + }, + { + "cell_type": "markdown", + "id": "2c8f3268", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Run classifier performance test with multiple models\n", + "\n", + "Now, we'll use the `input_grid` to run the [`ClassifierPerformance` test](https://docs.validmind.ai/tests/model_validation/sklearn/ClassifierPerformance.html) on all four models using the testing dataset (`vm_test_ds`).\n", + "\n", + "When running individual tests, you can use a custom `result_id` to tag the individual result with a unique identifier by appending this `result_id` to the `test_id` with a `:` separator. We'll append an identifier to signify that this test was run on `all_models` to differentiate this test run from other runs:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2e48ce1e", + "metadata": {}, + "outputs": [], + "source": [ + "perf_comparison_result = vm.tests.run_test(\n", + " \"validmind.model_validation.sklearn.ClassifierPerformance:all_models\",\n", + " input_grid={\n", + " \"dataset\": [vm_test_ds],\n", + " \"model\": [vm_model_xgb, vm_model_rf, vm_model_lr, vm_model_dt],\n", + " },\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "81cbf144", + "metadata": {}, + "source": [ + "Our output indicates that the XGBoost and random forest classification models provide the strongest overall classification performance, so we'll continue our testing with those two models as input only." + ] + }, + { + "cell_type": "markdown", + "id": "3d3fb6ec", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Run classifier performance test with multiple parameter values\n", + "\n", + "Next, let's run the classifier performance test with the `param_grid` object, which runs the same test multiple times with different parameter values. We'll append an identifier to signify that this test was run with our `parameter_grid` configuration:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d0ad94c9", + "metadata": {}, + "outputs": [], + "source": [ + "parameter_comparison_result = vm.tests.run_test(\n", + " \"validmind.model_validation.sklearn.ClassifierPerformance:parameter_grid\",\n", + " input_grid={\n", + " \"dataset\": [vm_test_ds],\n", + " \"model\": [vm_model_xgb,vm_model_rf]\n", + " },\n", + " param_grid={\n", + " \"average\": [\"macro\", \"micro\"]\n", + " },\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "508c7546", + "metadata": {}, + "source": [ + "\n", + "\n", + "#### Run comparison test with multiple datasets\n", + "\n", + "Let's also run the [ROCCurve test](https://docs.validmind.ai/tests/model_validation/sklearn/ROCCurve.html) using `input_grid` to iterate through multiple datasets, which plots the ROC curves for the training (`vm_train_ds`) and test (`vm_test_ds`) datasets side by side — a common scenario when you want to compare the performance of a model on the training and test datasets and visually assess how much performance is lost in the test dataset.\n", + "\n", + "We'll also need to assign predictions to the training dataset for the random forest classifier model, since we didn't do that in our earlier setup:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "96c3b426", + "metadata": {}, + "outputs": [], + "source": [ + "vm_train_ds.assign_predictions(model=vm_model_rf)" + ] + }, + { + "cell_type": "markdown", + "id": "2be82bae", + "metadata": {}, + "source": [ + "We'll append an identifier to signify that this test was run with our `train_vs_test` dataset comparison configuration:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4056aa1e", + "metadata": {}, + "outputs": [], + "source": [ + "roc_curve_result = vm.tests.run_test(\n", + " \"validmind.model_validation.sklearn.ROCCurve:train_vs_test\",\n", + " input_grid={\n", + " \"dataset\": [vm_train_ds, vm_test_ds],\n", + " \"model\": [vm_model_xgb,vm_model_rf],\n", + " },\n", + ")" + ] + }, + { + "cell_type": "markdown", + "id": "a05570d5", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Work with test results\n", + "\n", + "Every test result returned by the `run_test()` function has a [`.log()` method](https://docs.validmind.ai/validmind/validmind/vm_models.html#TestResult.log) that can be used to send the test results to the ValidMind Platform. When logging individual test results to the platform, you'll need to manually add those results to the desired section of the model documentation.\n", + "\n", + "You can do this through the ValidMind Platform interface after logging your test results ([Learn more ...](https://docs.validmind.ai/developer/model-documentation/work-with-test-results.html)), or directly via the ValidMind Library when calling `.log()` by providing an optional `section_id`. The `section_id` should be a string that matches the title of a section in the documentation template in `snake_case`.\n", + "\n", + "Let's log the results of the classifier performance test (`perf_comparison_result`) and the ROCCurve (`roc_curve_result`) test in the `model_evaluation` section of the documentation — present in the template we previewed in the beginning of this notebook:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e119bf1e", + "metadata": {}, + "outputs": [], + "source": [ + "perf_comparison_result.log(section_id=\"model_evaluation\")\n", + "roc_curve_result.log(section_id=\"model_evaluation\")" + ] + }, + { + "cell_type": "markdown", + "id": "ab5205ee", + "metadata": {}, + "source": [ + "Finally, let's head to the model we connected to at the beginning of this notebook and view our inserted test results in the updated documentation ([Need more help?](https://docs.validmind.ai/guide/model-documentation/working-with-model-documentation.html)):\n", + "\n", + "1. From the **Inventory** in the ValidMind Platform, go to the model you connected to earlier.\n", + "\n", + "2. In the left sidebar that appears for your model, click **Development** under Documents.\n", + "\n", + "3. Expand the **3.2. Model Evaluation** section.\n", + "\n", + "4. Confirm that `perf_comparison_result` and `roc_curve_result` display in this section as expected." + ] + }, + { + "cell_type": "markdown", + "id": "eb196aac", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Next steps\n", + "\n", + "Now that you know how to run comparison tests with the ValidMind Library, you’re ready to take the next step. Extend the functionality of `run_test()` with your own custom test functions that can be incorporated into documentation templates just like any default out-of-the-box ValidMind test.\n", + "\n", + "
Learn how to implement custom tests with the ValidMind Library.\n", + "

\n", + "Check out our Implement comparison tests notebook for code examples and usage of key functions.
" + ] + }, + { + "cell_type": "markdown", + "id": "083c1d8d", + "metadata": {}, + "source": [ + "\n", + "\n", + "### Discover more learning resources\n", + "\n", + "We offer many interactive notebooks to help you automate testing, documenting, validating, and more:\n", + "\n", + "- [Run tests & test suites](https://docs.validmind.ai/developer/how-to/testing-overview.html)\n", + "- [Use ValidMind Library features](https://docs.validmind.ai/developer/how-to/feature-overview.html)\n", + "- [Code samples by use case](https://docs.validmind.ai/guide/samples-jupyter-notebooks.html)\n", + "\n", + "Or, visit our [documentation](https://docs.validmind.ai/) to learn more about ValidMind." + ] + }, + { + "cell_type": "markdown", + "id": "efba0f57", + "metadata": {}, + "source": [ + "\n", + "\n", + "## Upgrade ValidMind\n", + "\n", + "
After installing ValidMind, you’ll want to periodically make sure you are on the latest version to access any new features and other enhancements.
\n", + "\n", + "Retrieve the information for the currently installed version of ValidMind:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0d35972c", + "metadata": { + "vscode": { + "languageId": "plaintext" } + }, + "outputs": [], + "source": [ + "%pip show validmind" + ] + }, + { + "cell_type": "markdown", + "id": "abcd07ef", + "metadata": {}, + "source": [ + "If the version returned is lower than the version indicated in our [production open-source code](https://github.com/validmind/validmind-library/blob/prod/validmind/__version__.py), restart your notebook and run:\n", + "\n", + "```bash\n", + "%pip install --upgrade validmind\n", + "```" + ] + }, + { + "cell_type": "markdown", + "id": "5fe70b90", + "metadata": {}, + "source": [ + "You may need to restart your kernel after running the upgrade package for changes to be applied." + ] + }, + { + "cell_type": "markdown", + "id": "copyright-54faffd51a5a4717a02b6be426d6b441", + "metadata": {}, + "source": [ + "\n", + "\n", + "\n", + "\n", + "***\n", + "\n", + "Copyright © 2023-2026 ValidMind Inc. All rights reserved.
\n", + "Refer to [LICENSE](https://github.com/validmind/validmind-library/blob/main/LICENSE) for details.
\n", + "SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" }, - "nbformat": 4, - "nbformat_minor": 5 + "language_info": { + "name": "python", + "version": "3.10" + } + }, + "nbformat": 4, + "nbformat_minor": 5 } diff --git a/site/notebooks/how_to/tests/run_tests/configure_tests/enable_pii_detection.ipynb b/site/notebooks/how_to/tests/run_tests/configure_tests/enable_pii_detection.ipynb index e0ae28fff7..ddbd734e9d 100644 --- a/site/notebooks/how_to/tests/run_tests/configure_tests/enable_pii_detection.ipynb +++ b/site/notebooks/how_to/tests/run_tests/configure_tests/enable_pii_detection.ipynb @@ -157,7 +157,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To use PII detection powered by [Microsoft Presidio](https://microsoft.github.io/presidio/), install the library with the explicit `[pii-detection]` extra specifier:" ] diff --git a/site/notebooks/quickstart/quickstart_model_documentation.Rmd b/site/notebooks/quickstart/quickstart_model_documentation.Rmd new file mode 100644 index 0000000000..fc48d8c445 --- /dev/null +++ b/site/notebooks/quickstart/quickstart_model_documentation.Rmd @@ -0,0 +1,227 @@ +--- +title: "Quickstart for Model Documentation (R)" +author: "ValidMind" +date: "2026-03-18" +output: html_document +--- + +# Quickstart for Model Documentation + +Learn the basics of using ValidMind to document models as part of a model development workflow using R. This notebook uses the ValidMind R package (a `reticulate` wrapper around the Python library) to generate a draft of documentation for a binary classification model. + +We will: + +1. Import a sample dataset and preprocess it +2. Split the datasets and initialize them for use with ValidMind +3. Train a logistic regression (GLM) model and initialize it for use with testing +4. Run the full suite of documentation tests, sending results to the ValidMind Platform + +## Setting up + +The Python path is auto-configured via the `VALIDMIND_PYTHON` environment variable. +If not set, it falls back to the system Python. For local development, create a +`.Renviron` file in the project root with `VALIDMIND_PYTHON=.venv/bin/python`. + +```{r setup, include=FALSE} +library(reticulate) + +python_version <- Sys.getenv("VALIDMIND_PYTHON", Sys.which("python")) +if (nchar(python_version) > 0 && !startsWith(python_version, "/")) { + python_version <- file.path(getwd(), python_version) +} +use_python(python_version, required = TRUE) + +library(validmind) +library(dplyr) +library(caTools) +library(knitr) + +knitr::opts_chunk$set(warning = FALSE, message = FALSE) +``` + +## Initialize the ValidMind Library + +Log in to the [ValidMind Platform](https://app.prod.validmind.ai) and register a model: + +1. Navigate to **Inventory** and click **+ Register Model**. +2. Under **Documents > Development**, select the `Binary classification` template. +3. Go to **Getting Started**, select `Development` from the **DOCUMENT** drop-down, and copy the code snippet. + +Replace the placeholder values below with your own credentials: + +```{r} +vm_r <- vm( + api_host = "https://app.prod.validmind.ai/api/v1/tracking", + api_key = "", + api_secret = "", + model = "", + document = "documentation" +) +``` + +## Preview the documentation template + +Verify the connection and see the documentation structure: + +```{r} +py_print(vm_r$preview_template()) +``` + +## Load the demo dataset + +We use the Bank Customer Churn dataset for this demonstration: + +```{r} +customer_churn <- reticulate::import( + "validmind.datasets.classification.customer_churn" +) + +cat(sprintf( + paste0( + "Loaded demo dataset with:\n\n\t- Target column: '%s'", + "\n\t- Class labels: %s\n" + ), + customer_churn$target_column, + paste( + names(customer_churn$class_labels), + customer_churn$class_labels, + sep = ": ", collapse = ", " + ) +)) + +data <- customer_churn$load_data() +head(data) +``` + +## Initialize the raw dataset + +Before running tests, initialize a ValidMind dataset object for the raw data: + +```{r} +vm_raw_dataset <- vm_r$init_dataset( + dataset = data, + input_id = "raw_dataset", + target_column = customer_churn$target_column, + class_labels = customer_churn$class_labels +) +``` + +## Preprocess the raw dataset + +Handle categorical variables using one-hot encoding and remove unnecessary columns: + +```{r} +# load_data() already drops RowNumber, CustomerId, Surname +# One-hot encode categorical variables +geo_dummies <- model.matrix(~ Geography - 1, data = data) +gender_dummies <- model.matrix(~ Gender - 1, data = data) +data_processed <- data %>% select(-Geography, -Gender) +data_processed <- cbind(data_processed, geo_dummies, gender_dummies) +``` + +### Split the dataset + +Split into training (60%), validation (20%), and test (20%) sets: + +```{r} +set.seed(42) + +# First split: 80% train+validation, 20% test +target_col <- customer_churn$target_column +split1 <- sample.split(data_processed[[target_col]], SplitRatio = 0.8) +train_val_data <- subset(data_processed, split1 == TRUE) +test_data <- subset(data_processed, split1 == FALSE) + +# Second split: 75% train, 25% validation (of the 80% = 60/20 overall) +split2 <- sample.split(train_val_data[[target_col]], SplitRatio = 0.75) +train_data <- subset(train_val_data, split2 == TRUE) +validation_data <- subset(train_val_data, split2 == FALSE) +``` + +## Train a logistic regression model + +Train a GLM with a binomial family (logistic regression): + +```{r} +formula <- as.formula(paste(target_col, "~ .")) +model <- glm(formula, data = train_data, family = binomial) +summary(model) +``` + +## Initialize the ValidMind datasets + +```{r} +vm_train_ds <- vm_r$init_dataset( + dataset = train_data, + input_id = "train_dataset", + target_column = customer_churn$target_column +) + +vm_test_ds <- vm_r$init_dataset( + dataset = test_data, + input_id = "test_dataset", + target_column = customer_churn$target_column +) +``` + +## Initialize a model object + +Save the R model and initialize it with ValidMind: + +```{r} +model_path <- save_model(model) + +vm_model <- vm_r$init_r_model( + model_path = model_path, + input_id = "model" +) +``` + +### Assign predictions + +Link model predictions to the training and testing datasets: + +```{r} +vm_train_ds$assign_predictions(model = vm_model) +vm_test_ds$assign_predictions(model = vm_model) +``` + +## Run the full suite of tests + +Build the test configuration that maps each test to its required inputs: + +```{r} +# Import the test config helper from the Python customer_churn module +customer_churn <- reticulate::import( + "validmind.datasets.classification.customer_churn" +) +test_config <- customer_churn$get_demo_test_config() +``` + +Preview the test configuration: + +```{r} +vm_utils <- reticulate::import("validmind.utils") +py_print(vm_utils$preview_test_config(test_config)) +``` + +Run the full documentation test suite and upload results to the ValidMind Platform: + +```{r} +full_suite <- vm_r$run_documentation_tests(config = test_config) +``` + +## Next steps + +Head to the [ValidMind Platform](https://app.prod.validmind.ai) to view the generated documentation: + +1. Navigate to **Inventory** and select your model. +2. Click **Development** under Documents to see the full draft of your model documentation. + +From there, you can make qualitative edits, collaborate with validators, and submit for approval. + +--- + +*Copyright 2023-2026 ValidMind Inc. All rights reserved.* +*Refer to [LICENSE](https://github.com/validmind/validmind-library/blob/main/LICENSE) for details.* +*SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial* diff --git a/site/notebooks/quickstart/quickstart_model_documentation.ipynb b/site/notebooks/quickstart/quickstart_model_documentation.ipynb index e7c28cf339..40287aa57f 100644 --- a/site/notebooks/quickstart/quickstart_model_documentation.ipynb +++ b/site/notebooks/quickstart/quickstart_model_documentation.ipynb @@ -184,7 +184,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/quickstart/quickstart_model_validation.Rmd b/site/notebooks/quickstart/quickstart_model_validation.Rmd new file mode 100644 index 0000000000..547aaf29d5 --- /dev/null +++ b/site/notebooks/quickstart/quickstart_model_validation.Rmd @@ -0,0 +1,335 @@ +--- +title: "Quickstart for Model Validation (R)" +author: "ValidMind" +date: "2026-03-18" +output: html_document +--- + +# Quickstart for Model Validation + +Learn the basics of using ValidMind to validate models as part of a model validation workflow using R. This notebook uses the ValidMind R package (a `reticulate` wrapper around the Python library) to generate a draft of a validation report for a binary classification model. + +We will: + +1. Import a sample dataset and preprocess it, then initialize datasets for use with ValidMind +2. Independently verify data quality tests performed on datasets by model development +3. Train a champion model for evaluation +4. Run model evaluation tests with the ValidMind Library + +## Setting up + +The Python path is auto-configured via the `VALIDMIND_PYTHON` environment variable. +If not set, it falls back to the system Python. For local development, create a +`.Renviron` file in the project root with `VALIDMIND_PYTHON=.venv/bin/python`. + +```{r setup, include=FALSE} +library(reticulate) + +python_version <- Sys.getenv("VALIDMIND_PYTHON", Sys.which("python")) +if (nchar(python_version) > 0 && !startsWith(python_version, "/")) { + python_version <- file.path(getwd(), python_version) +} +use_python(python_version, required = TRUE) + +library(validmind) +library(dplyr) +library(caTools) +library(knitr) + +knitr::opts_chunk$set(warning = FALSE, message = FALSE) +``` + +## Initialize the ValidMind Library + +Log in to the [ValidMind Platform](https://app.prod.validmind.ai) and register a model: + +1. Navigate to **Inventory** and click **+ Register Model**. +2. Assign yourself as a **Validator** (remove yourself from Owner and Developer roles). +3. Under **Documents > Validation**, select the `Generic Validation Report` template. +4. Go to **Getting Started**, select `Validation` from the **DOCUMENT** drop-down, and copy the code snippet. + +Replace the placeholder values below with your own credentials: + +```{r} +vm_r <- vm( + api_host = "https://app.prod.validmind.ai/api/v1/tracking", + api_key = "", + api_secret = "", + model = "", + document = "validation-report" +) +``` + +## Preview the validation report template + +Verify the connection and see the validation report structure: + +```{r} +py_print(vm_r$preview_template()) +``` + +## Identify available tests + +List the tasks and tags available in the ValidMind test library: + +```{r} +vm_r$tests$list_tasks_and_tags() +``` + +List all data quality tests for classification: + +```{r} +vm_r$tests$list_tests(tags = list("data_quality"), task = "classification") +``` + +## Load the demo dataset + +We use the Bank Customer Churn dataset for this demonstration: + +```{r} +customer_churn <- reticulate::import( + "validmind.datasets.classification.customer_churn" +) + +cat(sprintf( + paste0( + "Loaded demo dataset with:\n\n\t- Target column: '%s'", + "\n\t- Class labels: %s\n" + ), + customer_churn$target_column, + paste( + names(customer_churn$class_labels), + customer_churn$class_labels, + sep = ": ", collapse = ", " + ) +)) + +data <- customer_churn$load_data() +head(data) +``` + +## Preprocess the raw dataset + +Handle categorical variables using one-hot encoding and remove unnecessary columns: + +```{r} +# load_data() already drops RowNumber, CustomerId, Surname +# One-hot encode categorical variables +geo_dummies <- model.matrix(~ Geography - 1, data = data) +gender_dummies <- model.matrix(~ Gender - 1, data = data) +data_processed <- data %>% select(-Geography, -Gender) +data_processed <- cbind(data_processed, geo_dummies, gender_dummies) +``` + +### Split the dataset + +Split into training (60%), validation (20%), and test (20%) sets: + +```{r} +set.seed(42) + +# First split: 80% train+validation, 20% test +target_col <- customer_churn$target_column +split1 <- sample.split(data_processed[[target_col]], SplitRatio = 0.8) +train_val_data <- subset(data_processed, split1 == TRUE) +test_data <- subset(data_processed, split1 == FALSE) + +# Second split: 75% train, 25% validation (of the 80% = 60/20 overall) +split2 <- sample.split(train_val_data[[target_col]], SplitRatio = 0.75) +train_data <- subset(train_val_data, split2 == TRUE) +validation_data <- subset(train_val_data, split2 == FALSE) +``` + +### Separate features and targets + +```{r} +x_train <- train_data %>% select(-all_of(target_col)) +y_train <- train_data[[target_col]] +``` + +## Initialize the ValidMind datasets + +```{r} +vm_raw_dataset <- vm_r$init_dataset( + dataset = data, + input_id = "raw_dataset", + target_column = customer_churn$target_column, + class_labels = customer_churn$class_labels +) + +vm_train_ds <- vm_r$init_dataset( + dataset = train_data, + input_id = "train_dataset", + target_column = customer_churn$target_column +) + +vm_validation_ds <- vm_r$init_dataset( + dataset = validation_data, + input_id = "validation_dataset", + target_column = customer_churn$target_column +) + +vm_test_ds <- vm_r$init_dataset( + dataset = test_data, + input_id = "test_dataset", + target_column = customer_churn$target_column +) +``` + +## Run data quality tests + +### Run an individual data quality test + +Run the ClassImbalance test on the raw dataset and log it to the platform: + +```{r} +vm_r$tests$run_test( + test_id = "validmind.data_validation.ClassImbalance", + inputs = list(dataset = vm_raw_dataset) +)$log() +``` + +### Run data comparison tests + +Compare class imbalance across dataset splits: + +```{r} +comparison_tests <- list( + "validmind.data_validation.ClassImbalance:train_vs_validation" = list( + input_grid = list(dataset = list("train_dataset", "validation_dataset")) + ), + "validmind.data_validation.ClassImbalance:train_vs_test" = list( + input_grid = list(dataset = list("train_dataset", "test_dataset")) + ) +) + +for (test_name in names(comparison_tests)) { + cat(paste0("Running: ", test_name, "\n")) + config <- comparison_tests[[test_name]] + tryCatch({ + vm_r$tests$run_test( + test_name, + input_grid = config$input_grid + )$log() + }, error = function(e) { + cat(paste0("Error running test ", test_name, ": ", e$message, "\n")) + }) +} +``` + +## Train the champion model + +Train a logistic regression (GLM) to serve as the champion model: + +```{r} +formula <- as.formula(paste(target_col, "~ .")) +model <- glm(formula, data = train_data, family = binomial) +summary(model) +``` + +## Initialize the model object + +Save the R model and initialize it with ValidMind: + +```{r} +model_path <- save_model(model) + +vm_xgboost <- vm_r$init_r_model( + model_path = model_path, + input_id = "xgboost_champion" +) +``` + +### Assign predictions + +Link model predictions to the training and testing datasets: + +```{r} +vm_train_ds$assign_predictions(model = vm_xgboost) +vm_test_ds$assign_predictions(model = vm_xgboost) +``` + +## Run model evaluation tests + +### Run model performance tests + +List available model performance tests: + +```{r} +vm_r$tests$list_tests(tags = list("model_performance"), task = "classification") +``` + +Run and log performance tests: + +```{r} +performance_tests <- c( + "validmind.model_validation.sklearn.ClassifierPerformance:xgboost_champion", + "validmind.model_validation.sklearn.ConfusionMatrix:xgboost_champion", + "validmind.model_validation.sklearn.ROCCurve:xgboost_champion" +) + +for (test in performance_tests) { + cat(paste0("Running: ", test, "\n")) + vm_r$tests$run_test( + test, + inputs = list(dataset = vm_test_ds, model = vm_xgboost) + )$log() +} +``` + +### Run diagnostic tests + +Assess the model for overfitting: + +```{r} +vm_r$tests$run_test( + test_id = paste0( + "validmind.model_validation.sklearn.OverfitDiagnosis", + ":xgboost_champion" + ), + input_grid = list( + datasets = list(list(vm_train_ds, vm_test_ds)), + model = list(vm_xgboost) + ) +)$log() +``` + +Test robustness: + +```{r} +vm_r$tests$run_test( + test_id = paste0( + "validmind.model_validation.sklearn.RobustnessDiagnosis", + ":xgboost_champion" + ), + input_grid = list( + datasets = list(list(vm_train_ds, vm_test_ds)), + model = list(vm_xgboost) + ) +)$log() +``` + +### Run feature importance tests + +Note: `PermutationFeatureImportance` and `SHAPGlobalImportance` are not supported for R models. + +```{r} +vm_r$tests$run_test( + "validmind.model_validation.FeaturesAUC:xgboost_champion", + inputs = list(dataset = vm_test_ds, model = vm_xgboost) +)$log() +``` + +## Next steps + +Head to the [ValidMind Platform](https://app.prod.validmind.ai) to view the validation report: + +1. Navigate to **Inventory** and select your model. +2. Click **Validation** under Documents. +3. Include your logged test results as evidence, create risk assessment notes, and assess compliance. + +--- + +*Copyright 2023-2026 ValidMind Inc. All rights reserved.* +*Refer to [LICENSE](https://github.com/validmind/validmind-library/blob/main/LICENSE) for details.* +*SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial* diff --git a/site/notebooks/quickstart/quickstart_model_validation.ipynb b/site/notebooks/quickstart/quickstart_model_validation.ipynb index 640e64015b..63e17f2a88 100644 --- a/site/notebooks/quickstart/quickstart_model_validation.ipynb +++ b/site/notebooks/quickstart/quickstart_model_validation.ipynb @@ -259,7 +259,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/tutorials/model_development/1-set_up_validmind.ipynb b/site/notebooks/tutorials/model_development/1-set_up_validmind.ipynb index f82f57eaa1..4244924b9a 100644 --- a/site/notebooks/tutorials/model_development/1-set_up_validmind.ipynb +++ b/site/notebooks/tutorials/model_development/1-set_up_validmind.ipynb @@ -171,7 +171,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/tutorials/model_validation/1-set_up_validmind_for_validation.ipynb b/site/notebooks/tutorials/model_validation/1-set_up_validmind_for_validation.ipynb index c5dc1fb394..05ad11c2c3 100644 --- a/site/notebooks/tutorials/model_validation/1-set_up_validmind_for_validation.ipynb +++ b/site/notebooks/tutorials/model_validation/1-set_up_validmind_for_validation.ipynb @@ -261,7 +261,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/use_cases/agents/document_agentic_ai.ipynb b/site/notebooks/use_cases/agents/document_agentic_ai.ipynb index 89e8152214..3c3b6817b1 100644 --- a/site/notebooks/use_cases/agents/document_agentic_ai.ipynb +++ b/site/notebooks/use_cases/agents/document_agentic_ai.ipynb @@ -194,7 +194,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.9 <= x <= 3.11
\n", + "Python 3.9 <= x <= 3.14\n", "\n", "Let's begin by installing the ValidMind Library with large language model (LLM) support:" ] diff --git a/site/notebooks/use_cases/model_validation/validate_application_scorecard.ipynb b/site/notebooks/use_cases/model_validation/validate_application_scorecard.ipynb index f3df8a617e..7857d42e0a 100644 --- a/site/notebooks/use_cases/model_validation/validate_application_scorecard.ipynb +++ b/site/notebooks/use_cases/model_validation/validate_application_scorecard.ipynb @@ -247,7 +247,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/notebooks/use_cases/nlp_and_llm/rag_benchmark_demo.ipynb b/site/notebooks/use_cases/nlp_and_llm/rag_benchmark_demo.ipynb index 1e3eb07b67..1b56fa1b07 100644 --- a/site/notebooks/use_cases/nlp_and_llm/rag_benchmark_demo.ipynb +++ b/site/notebooks/use_cases/nlp_and_llm/rag_benchmark_demo.ipynb @@ -159,7 +159,7 @@ "\n", "
Recommended Python versions\n", "

\n", - "Python 3.8 <= x <= 3.11
\n", + "Python 3.8 <= x <= 3.14\n", "\n", "To install the library:" ] diff --git a/site/releases/2025/2025-mar-07/release-notes.qmd b/site/releases/2025/2025-mar-07/release-notes.qmd index f2a363f65b..1e76360c0c 100644 --- a/site/releases/2025/2025-mar-07/release-notes.qmd +++ b/site/releases/2025/2025-mar-07/release-notes.qmd @@ -58,7 +58,7 @@ We've introduced an interactive historical record of breaking changes and deprec - These histories grouped by calendar year are searchable and can be filtered by product area ({{< var validmind.developer >}}, {{< var validmind.platform >}}), associated versioning, and the type of change. ::: {.column-margin} -[Breaking changes & deprecations](/releases/breaking-changes/breaking-changes.qmd){.button .button-green target="_blank"} +[Breaking changes & deprecations](/releases/breaking-changes-and-deprecations.qmd){.button .button-green target="_blank"} ::: From 31541c21c3bc9256d75faafe34023ddbb7f28cee Mon Sep 17 00:00:00 2001 From: Nik Richers Date: Thu, 2 Apr 2026 17:58:30 -0700 Subject: [PATCH 10/10] Nrichers/sc 15354/workflows fix notebook execution workflows (#1263) * Allow check-in of baseline template schema docs and make CI workflow fail harder if necessary * Fix README.md * Exclude notebooks/EXECUTED/* from AWS S3 sync * Another test notebook * Move notebook into EXECUTED path * Also --exclude executed notebooks on staging & prod * Delete another test notebook * Update scripts/generate_template_schema_docs.py --- .github/workflows/deploy-docs-prod.yaml | 2 +- .github/workflows/deploy-docs-staging.yaml | 2 +- .github/workflows/validate-docs-site.yaml | 2 +- .gitignore | 3 - README.md | 4 +- scripts/generate_template_schema_docs.py | 56 +- site/Makefile | 2 +- .../templates/_template-schema-generated.qmd | 1143 +++++++++++++++++ .../customize-document-templates.qmd | 7 - 9 files changed, 1191 insertions(+), 30 deletions(-) create mode 100644 site/guide/templates/_template-schema-generated.qmd diff --git a/.github/workflows/deploy-docs-prod.yaml b/.github/workflows/deploy-docs-prod.yaml index c2e62fb39b..24697c2f30 100644 --- a/.github/workflows/deploy-docs-prod.yaml +++ b/.github/workflows/deploy-docs-prod.yaml @@ -117,7 +117,7 @@ jobs: run: aws configure set aws_access_key_id ${{ secrets.AWS_ACCESS_KEY_ID_PROD }} && aws configure set aws_secret_access_key ${{ secrets.AWS_SECRET_ACCESS_KEY_PROD }} && aws configure set default.region us-east-1 - name: Deploy docs prod site - run: aws s3 sync site/_site s3://validmind-docs-prod/site --delete --exclude "installation/omnibus/*" --exclude "installation/helm-repo/*" --exclude "llm/*" && aws cloudfront create-invalidation --distribution-id E2BGG3USKQTR9W --paths "/*" --no-cli-pager + run: aws s3 sync site/_site s3://validmind-docs-prod/site --delete --exclude "installation/omnibus/*" --exclude "installation/helm-repo/*" --exclude "notebooks/EXECUTED/*" --exclude "llm/*" && aws cloudfront create-invalidation --distribution-id E2BGG3USKQTR9W --paths "/*" --no-cli-pager # Release headroom and shrink before final lightweight steps & post-job - name: Release reserve & shrink diff --git a/.github/workflows/deploy-docs-staging.yaml b/.github/workflows/deploy-docs-staging.yaml index 48e76d348f..4c8f8fb69b 100644 --- a/.github/workflows/deploy-docs-staging.yaml +++ b/.github/workflows/deploy-docs-staging.yaml @@ -120,7 +120,7 @@ jobs: run: aws configure set aws_access_key_id ${{ secrets.AWS_ACCESS_KEY_ID_STAGING }} && aws configure set aws_secret_access_key ${{ secrets.AWS_SECRET_ACCESS_KEY_STAGING }} && aws configure set default.region us-west-2 - name: Deploy docs staging site - run: aws s3 sync site/_site s3://validmind-docs-staging/site --delete --exclude "installation/helm-repo/*" --exclude "pr_previews/*" --exclude "llm/*" && aws cloudfront create-invalidation --distribution-id ESWVTZYFL873V --paths "/*" --no-cli-pager + run: aws s3 sync site/_site s3://validmind-docs-staging/site --delete --exclude "installation/helm-repo/*" --exclude "pr_previews/*" --exclude "notebooks/EXECUTED/*" --exclude "llm/*" && aws cloudfront create-invalidation --distribution-id ESWVTZYFL873V --paths "/*" --no-cli-pager # Release headroom and shrink before final lightweight steps & post-job - name: Release reserve & shrink diff --git a/.github/workflows/validate-docs-site.yaml b/.github/workflows/validate-docs-site.yaml index da1d3bc9d1..309a7b302f 100644 --- a/.github/workflows/validate-docs-site.yaml +++ b/.github/workflows/validate-docs-site.yaml @@ -130,7 +130,7 @@ jobs: run: aws configure set aws_access_key_id ${{ secrets.AWS_ACCESS_KEY_ID_STAGING }} && aws configure set aws_secret_access_key ${{ secrets.AWS_SECRET_ACCESS_KEY_STAGING }} && aws configure set default.region us-east-1 - name: Deploy PR preview - run: aws s3 sync site/_site s3://validmind-docs-staging/site/pr_previews/${{ github.head_ref }} --delete && aws cloudfront create-invalidation --distribution-id ESWVTZYFL873V --paths "/*" --no-cli-pager + run: aws s3 sync site/_site s3://validmind-docs-staging/site/pr_previews/${{ github.head_ref }} --delete --exclude "notebooks/EXECUTED/*" && aws cloudfront create-invalidation --distribution-id ESWVTZYFL873V --paths "/*" --no-cli-pager - name: Post comment with preview URL uses: actions/github-script@v6 diff --git a/.gitignore b/.gitignore index 9684da543c..f4de6a8af5 100644 --- a/.gitignore +++ b/.gitignore @@ -36,9 +36,6 @@ site/validmind-docs.yaml # Python API docs are now generated on the fly site/validmind -# Generated template schema documentation -site/guide/templates/_template-schema-generated.qmd - # Cursor rules .cursor/rules/ .cursor/skills/ diff --git a/README.md b/README.md index 1734a7b31e..2c188de20c 100644 --- a/README.md +++ b/README.md @@ -150,7 +150,7 @@ Some documentation content is auto-generated from backend source files. These sc #### Template schema documentation -The template schema reference in `site/guide/templates/customize-document-templates.qmd` is auto-generated from the backend JSON Schema. CI workflows generate this automatically, but you can also regenerate locally: +The template schema reference in `site/guide/templates/customize-document-templates.qmd` is auto-generated from the backend JSON Schema, overwriting any baseline output checked into this repo. You can also regenerate locally and commit: ```bash cd site @@ -172,7 +172,7 @@ make template-schema-docs The script reads from: - `backend/src/backend/templates/documentation/model_documentation/mdd_template_schema_v5.json` — template schema definition -Output: `site/guide/templates/_template-schema-generated.qmd` +Output: Content is injected directly into `site/guide/templates/customize-document-templates.qmd` between marker comments. #### Stylesheet organization (IN PROGRESS) diff --git a/scripts/generate_template_schema_docs.py b/scripts/generate_template_schema_docs.py index 3becbac57d..c790489403 100644 --- a/scripts/generate_template_schema_docs.py +++ b/scripts/generate_template_schema_docs.py @@ -30,7 +30,10 @@ BACKEND_ROOT = Path(os.environ.get("BACKEND_ROOT", REPO_ROOT.parent / "backend")) SCHEMA_FILE = BACKEND_ROOT / "src/backend/templates/documentation/model_documentation/mdd_template_schema_v5_ui.json" -OUTPUT_FILE = REPO_ROOT / "site/guide/templates/_template-schema-generated.qmd" +TARGET_FILE = REPO_ROOT / "site/guide/templates/_template-schema-generated.qmd" + +# Minimum expected file size in bytes (sanity check for valid output) +MIN_OUTPUT_SIZE = 40000 # CSS stylesheets to link in the generated HTML STYLESHEETS = [ @@ -62,7 +65,7 @@ def main(): print(f"Generating schema documentation from {SCHEMA_FILE}") # Create temporary output - temp_output = OUTPUT_FILE.with_suffix(".tmp.html") + temp_output = TARGET_FILE.with_suffix(".tmp.html") subprocess.run([ "generate-schema-doc", @@ -98,6 +101,20 @@ def main(): '', html_content ) + # Strip the title tag (not needed when embedded in Quarto page) + html_content = re.sub( + r'\s*[^<]*\s*', + '\n', + html_content + ) + # Strip h1 headings (the page already has a heading from Quarto) + html_content = re.sub( + r'

[^<]*

', + '', + html_content + ) + # Clean up multiple consecutive blank lines + html_content = re.sub(r'\n{3,}', '\n\n', html_content) # Build stylesheet link tags stylesheet_links = '\n'.join( @@ -124,34 +141,45 @@ def main(): '' ) - # Copyright header to place before the raw HTML block - copyright_header = """""" - - # Wrap HTML in Quarto raw HTML block for .qmd file, with comment before - qmd_content = f"""{copyright_header} +Source: {SCHEMA_FILE.relative_to(BACKEND_ROOT.parent)} +--> ```{{=html}} {html_content} ``` """ - # Write final output - OUTPUT_FILE.parent.mkdir(parents=True, exist_ok=True) - with open(OUTPUT_FILE, "w") as f: - f.write(qmd_content) + # Validate output before writing + if len(output_content) < MIN_OUTPUT_SIZE: + print(f"Error: Generated output is too small ({len(output_content)} bytes)") + print("This likely indicates a generation failure.") + temp_output.unlink(missing_ok=True) + sys.exit(1) + + if "" not in html_content: + print("Error: Generated output does not contain valid HTML structure") + temp_output.unlink(missing_ok=True) + sys.exit(1) + + # Write to target file + TARGET_FILE.write_text(output_content) # Clean up temp file temp_output.unlink() - print(f"Generated {OUTPUT_FILE}") + print(f"Generated template schema documentation: {TARGET_FILE}") + print(f"Output size: {len(output_content)} bytes") if __name__ == "__main__": diff --git a/site/Makefile b/site/Makefile index 6773da52d5..d3372f286c 100644 --- a/site/Makefile +++ b/site/Makefile @@ -427,7 +427,7 @@ release-notes: template-schema-docs: @echo "\nGenerating template schema documentation ..." @if [ ! -d "$(SRC_ROOT)/backend" ]; then echo "Error: Backend not cloned. Run 'make clone' first."; exit 1; fi - @pip install -q json-schema-for-humans + @python -m pip install -q json-schema-for-humans @BACKEND_ROOT=$(SRC_ROOT)/backend python ../scripts/generate_template_schema_docs.py test-descriptions: diff --git a/site/guide/templates/_template-schema-generated.qmd b/site/guide/templates/_template-schema-generated.qmd new file mode 100644 index 0000000000..8edc74c349 --- /dev/null +++ b/site/guide/templates/_template-schema-generated.qmd @@ -0,0 +1,1143 @@ + + +```{=html} + + + + + + + + + + + + + + +
+ + +
+ +
Type: array
+

Top-level documentation sections shown in the template editor.

+ + + + + + + No Additional Items

Each item of this array must be:

+
+
+ + +
root + + + + section
Type: object
+

A section in the document tree.

+ + + No Additional Properties + + + + + + +
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + id
Type: string
+

Unique identifier for the section.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + title
Type: string
+

Title of the section.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + description
Type: string
+

Description of the section.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + order
Type: integer
+

Order of the section in the navigation menu. By default sections are ordered alphabetically. If order is specified, sections will be ordered by the order value, and then alphabetically.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + default_text
Type: string
+

Default text for the section. If set, a metadata content row will be created with this text when installing the template on a given project.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + index_only
Type: boolean
+

If true, the section will be displayed in the navigation menu, but it will not be accessible via direct link.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + condensed
Type: boolean
+

If true, the section will condense all of its subsections into a single section.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + guidelines
Type: array of string
+

Documentation or validation guidelines for the section.

+ + + + + + + No Additional Items

Each item of this array must be:

+
+
+ + +
root + + + + root items + + + + guidelines + + + + guidelines items
Type: string
+ + + + + + + +
+
+
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents
Type: array
+

Contents to be displayed on the section.

+ + + + + + + No Additional Items

Each item of this array must be:

+
+
+ + +
root + + + + root items + + + + contents + + + + section_contents
Type: object
+

Single content block rendered within a section.

+ + + + + + + + + +
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents + + + + contents items + + + + content_type
Type: enum (of string) Default: "text"
+
+

Must be one of:

+
  • "text"
  • "metadata_text"
  • "dynamic"
  • "metric"
  • "test"
  • "guideline"
  • "assessment_summary"
  • "unit_metric"
+
+ + + + + +
+
Examples:
+
"text"
+
+
"test"
+
+
+
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents + + + + contents items + + + + content_id

+

ID of the content to be displayed for the given content type.

+

+ +

+
+ + +
root + + + + root items + + + + contents + + + + contents items + + + + content_id + + + + oneOf + + + + item 0
Type: string
+ + + + + + +
+
Example:
+
"sample_text"
+
+
+
+ + +
root + + + + root items + + + + contents + + + + contents items + + + + content_id + + + + oneOf + + + + item 1
Type: array of string
+ + + + + + + No Additional Items

Each item of this array must be:

+
+
+ + +
root + + + + root items + + + + contents + + + + contents items + + + + content_id + + + + oneOf + + + + item 1 + + + + item 1 items
Type: string
+ + + + + + + +
+

+
Example:
+
[
+    "sample_text",
+    "section_intro"
+]
+
+
+
+ + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents + + + + contents items + + + + options
Type: object
+

Options for the content block.

+ + + + + + +
+
Examples:
+
{
+    "default_text": "This is a sample text block."
+}
+
+
{
+    "metric_id": "metric_1",
+    "title": "Custom Title for Metric 1"
+}
+
+
{
+    "test_id": "adf_test"
+}
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents + + + + contents items + + + + options + + + + default_text
Type: string
+

Default text for the content block. Only applicable for text content blocks.

+ + + + + + + +
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + contents + + + + contents items + + + + options + + + + title
Type: string
+

Title of the content block. Only applicable for metric and test content blocks.

+ + + + + + + +
+
+
+
+
+
+
+
+
+
+
+

+ +

+
+ +
+

Additional Properties of any type are allowed.

+ +
root + + + + root items + + + + contents + + + + contents items + + + + additionalProperties
Type: object
+ + + + + + + +
+
+
+
+
+
+
+
+
+
+
+
+
+

+ +

+
+ +
+
+ +
root + + + + root items + + + + sections
Type: array
+

Nested child sections.

+ + + + + + + No Additional Items

Each item of this array must be:

+
+
+ + +
root + + + + root items + + + + sections + + + + section
Type: object
+

A section in the document tree.

+Same definition as items +
+
+
+
+
+
+
+
+ +
+ +``` diff --git a/site/guide/templates/customize-document-templates.qmd b/site/guide/templates/customize-document-templates.qmd index 0d17104c7a..18622cbdd0 100644 --- a/site/guide/templates/customize-document-templates.qmd +++ b/site/guide/templates/customize-document-templates.qmd @@ -75,12 +75,6 @@ Once saved, your new template version becomes available for use. ### Template schema - - ::: {.column-page-inset-right} {{< include _template-schema-generated.qmd >}} @@ -173,4 +167,3 @@ c. Finish editing your template, then save a new version. [^9]: [Edit YAML templates](#edit-yaml-templates) [^10]: [Template schema](#template-schema) -