[26.04] Docs: ComfyUI release

CMakeLists.txt

@@ -0,0 +1 @@
+rocm_setup_version(VERSION 0.18.2)

amd-docs/.readthedocs.yaml

@@ -0,0 +1,21 @@
+# Read the Docs configuration file
+# See https://docs.readthedocs.io/en/stable/config-file/v2.html for details
+
+version: 2
+
+sphinx:
+   configuration: amd-docs/conf.py
+
+# RTD by default builds html only
+# Additional formats available for extra build time: htmlzip, pdf, epub
+formats: []
+
+python:
+   install:
+   - requirements: amd-docs/sphinx/requirements.txt
+
+# Defines build environment
+build:
+   os: ubuntu-24.04
+   tools:
+      python: "3.12"

amd-docs/about/license.rst

@@ -0,0 +1,12 @@
+.. meta::
+  :description: ComfyUI license
+  :keywords: ComfyUI, ROCm, developer, reference, agent
+
+.. _license:
+
+******************************************
+License
+******************************************
+
+.. include:: ../../LICENSE
+  :literal:

amd-docs/conf.py

@@ -0,0 +1,58 @@
+# Configuration file for the Sphinx documentation builder.
+#
+# This file only contains a selection of the most common options. For a full
+# list see the documentation:
+# https://www.sphinx-doc.org/en/master/usage/configuration.html
+
+import re
+
+'''
+html_theme is usually unchanged (rocm_docs_theme).
+flavor defines the site header display, select the flavor for the corresponding portals
+flavor options: rocm, rocm-docs-home, rocm-blogs, rocm-ds, instinct, ai-developer-hub, local, generic
+'''
+html_theme = "rocm_docs_theme"
+html_theme_options = {"flavor": "rocm-llmext", "repository_url": "https://github.com/AMD-Ecosystem/comfyui/"}
+
+'''
+docs_header_version is used to manually configure the version in the header. If
+there exists a non-null value mapped to docs_header_version, then the header in
+the documentation page will contain the given version string.
+'''
+html_context = {
+    "docs_header_version": "26.04"
+}
+
+
+# This section turns on/off article info
+setting_all_article_info = True
+all_article_info_os = ["linux"]
+all_article_info_author = ""
+
+# Dynamically extract component version
+with open('../CMakeLists.txt', encoding='utf-8') as f:
+    pattern = r'.*\brocm_setup_version\(VERSION\s+([0-9A-Za-z._-]+)' # Update according to each component's CMakeLists.txt
+    match = re.search(pattern,
+                      f.read())
+    if not match:
+        raise ValueError("VERSION not found!")
+    version_number = match[1]
+
+# for PDF output on Read the Docs
+project = "ComfyUI"
+author = "Advanced Micro Devices, Inc."
+copyright = "Copyright (c) 2026 Advanced Micro Devices, Inc. All rights reserved."
+version = version_number
+release = version_number
+
+external_toc_path = "./sphinx/_toc.yml" # Defines Table of Content structure definition path
+
+# Add more addtional package accordingly
+extensions = [
+    "rocm_docs", 
+    "sphinx.ext.autodoc", # for Python docstrings
+] 
+
+html_title = f"{project} {version_number} documentation"
+
+external_projects_current_project = "ComfyUI"

amd-docs/how-to/download-models.rst

@@ -0,0 +1,50 @@
+.. meta::
+  :description: Download models for ComfyUI
+  :keywords: ComfyUI, programming, agent, ROCm, example, sample, tutorial
+
+.. _comfyui-download-models:
+
+********************************************************************
+Download and use models in ComfyUI
+********************************************************************
+
+ComfyUI makes it easy to use models and build both simple and complex workflows with them.
+However, you must first download the models you want to use. While you can download models directly from the UI,
+the simplest and most stable method is to download them from a model repository (such as
+`Hugging Face <https://huggingface.co/models>`__ or `Civitai <https://civitai.com/models>`__)
+using command-line tools like ``curl``.
+
+When downloading models through the command line, ensure you save them to the directories expected by ComfyUI:
+
+* ``$COMFYUI_PATH/models/diffusion_models`` for diffusion models
+* ``$COMFYUI_PATH/models/unet`` for UNet models
+* ``$COMFYUI_PATH/models/vae`` for VAE models
+* ``$COMFYUI_PATH/models/text_encoders`` for text encoders
+* ``$COMFYUI_PATH/models/clip`` for CLIP models
+* ``$COMFYUI_PATH/models/checkpoints`` for model checkpoints
+
+Saving models to these directories makes it possible to use them in the standard ComfyUI nodes.
+
+Download custom nodes
+====================================================================
+
+ComfyUI comes with a rich library of nodes and templates. Additionally, a vast community builds custom nodes that you can import and install in two ways: from the command line or using the ComfyUI Manager.
+
+Command line installation
+--------------------------------------------------------------------
+
+To install nodes from the command line, clone a custom node repository into the ``custom_nodes`` directory and optionally install the dependencies:
+
+.. code-block:: bash
+
+   git clone <URL TO REPO> $COMFYUI_PATH/custom_nodes/<NAME OF NODE PACK>
+   pip install -r $COMFYUI_PATH/custom_nodes/<NAME OF NODE PACK>/requirements.txt
+
+.. note::
+
+   For newly installed nodes to appear in the UI, ensure you restart the server.
+
+ComfyUI Manager installation
+--------------------------------------------------------------------
+
+Installing nodes with the ComfyUI Manager is done directly in the UI. Select the custom nodes to install from the navigator and restart the server.

amd-docs/how-to/hunyuan3d-workflow.rst

@@ -0,0 +1,52 @@
+.. meta::
+  :description: ComfyUI example: Hunyuan3D 2.1 workflow
+  :keywords: ComfyUI, programming, agent, ROCm, example, sample, tutorial
+
+.. _run-comfyui-hunyuan-3d:
+
+********************************************************************
+Run the Hunyuan3D 2.1 template in ComfyUI
+********************************************************************
+
+`Hunyuan3D <https://3d.hunyuanglobal.com/>`__ is an open-source 3D asset generation model released by Tencent,
+capable of generating high-fidelity 3D models with high-resolution texture maps through text or images.
+
+Hunyuan3D 2.1 is Tencent's 3D asset generation system that turns single images into
+production-ready 3D models with physically-based rendering (PBR) materials. This release
+includes improved texture quality with finer surface details and enhanced three-dimensional depth perception.
+
+Setup
+====================================================================
+
+To run this workflow:
+
+1. Follow the :ref:`comfyui-on-rocm-installation` steps to set up the ComfyUI environment and launch the ComfyUI WebUI.
+
+2. Select **Template** from the navigation panel on the left and search for **Hunyuan 3D 2.1**.
+
+   .. image:: ../images/hunyuan3d-workflow-01.png
+      :alt: Template selection
+
+3. Double-click on **Hunyuan3D 2.1** template.
+
+4. Drag and drop an input image into the **Step 2 - Upload image here** node.
+
+   Example image: https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_image/hunyuan_image_example.png
+
+   .. image:: ../images/hunyuan3d-workflow-02.png
+      :alt: Upload image step
+
+5. Click **Run**.
+
+Modify resolution configuration
+--------------------------------------------------------------------
+
+To modify input resolution, edit the **resolution** option in the **Empty LaternHunyuan3Dv2** node.
+
+Additional resources
+--------------------------------------------------------------------
+
+- `ComfyUI Hunyuan3D-2 Examples <https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_image/>`__
+- `Hunyuan 3D Models <https://3d-models.hunyuan.tencent.com/>`__
+
+

amd-docs/how-to/wan-inference-workflow.rst

@@ -0,0 +1,56 @@
+.. meta::
+  :description: ComfyUI example: Wan 2.2 inference workflow
+  :keywords: ComfyUI, programming, agent, ROCm, example, sample, tutorial
+
+.. _run-comfyui-wan-inference:
+
+***************************************************************************
+Run the Wan 2.2 inference template in ComfyUI
+***************************************************************************
+
+`Wan 2.2 <https://docs.comfy.org/tutorials/video/wan/wan2_2>`__ is an advanced AI
+video generation model capable of producing high-resolution videos from text
+descriptions or input images. With support for various aspect ratios and
+customizable generation parameters, Wan 2.2 enables creative video synthesis
+for a wide range of applications.
+
+Setup
+====================================================================
+
+To run this workflow:
+
+1. Follow the :ref:`comfyui-on-rocm-installation` steps to set up the ComfyUI environment and launch the ComfyUI WebUI.
+
+2. Select **Template** from the navigation panel on the left and search for the **Wan 2.2 14B Text to Video** template.
+
+   .. figure:: ../images/wan22-templates-video.png
+      :align: center
+      :alt: ComfyUI Templates library with Video selected and the Wan 2.2 14B Text to Video card
+
+      **Templates → Video:** under *Generation type*, choose **Video**, then open the **Wan 2.2 14B Text to Video** template (for example, double-click the card).
+
+3. Once the template is loaded, click **Run** to execute the workflow.
+
+
+Trigger a run
+--------------------------------------------------------------------
+
+To trigger a run, enter prompts in the **CLIP Text Encode** nodes for positive and negative prompts, then click **Run**.
+
+.. figure:: ../images/wan22-workflow-canvas.png
+   :align: center
+   :alt: Wan 2.2 workflow graph with CLIP Text Encode nodes and the Run control
+
+   **Loaded workflow:** set positive and negative text in the **CLIP Text Encode** nodes, then use the blue **Run** button (top right).
+   Model loaders and other nodes follow the template layout (for example, **Wan2.2 T2V fp8_scaled**).
+
+You can see workflow progress at the top of the UI in the green progress bar near the top of the browser.
+
+
+Additional resources
+--------------------------------------------------------------------
+
+- `ComfyUI Wan 2.2 Documentation <https://docs.comfy.org/tutorials/video/wan/wan2_2>`__
+- `Wan 2.2 Model Files (HuggingFace) <https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged>`__
+
+

amd-docs/index.rst

@@ -0,0 +1,46 @@
+.. meta::
+  :description: ComfyUI documentation
+  :keywords: ComfyUI, ROCm, documentation, agent, GPU
+
+.. _comfyui-documentation-index:
+
+********************************************************************
+ComfyUI on ROCm documentation
+********************************************************************
+
+While you can build workflows for generative AI tasks purely in code, the growing interest in GenAI
+has led to increased demand for tools that don't require extensive programming knowledge.
+ComfyUI provides you with a simple drag-and-drop interface for building GenAI workflows.
+This guide will cover what ComfyUI is and how you can get it running on AMD Instinct GPUs.
+
+`ComfyUI <https://docs.comfy.org/index.html>`__ is an open-source,
+node-based interface for building and running image generation workflows with
+diffusion models such as Stable Diffusion. Its modular graph-based design lets
+you construct, customize, and share complex pipelines without writing code.
+
+ComfyUI is part of the `AMD LLM Extension toolkit
+<https://rocm.docs.amd.com/projects/rocm-llmext/en/docs-26.04/>`__.
+
+The ComfyUI public repository is located at `https://github.com/AMD-Ecosystem/ComfyUI/tree/release/0.18.2.amd0 <https://github.com/AMD-Ecosystem/ComfyUI/tree/release/0.18.2.amd0>`__.
+
+.. grid:: 2
+  :gutter: 3
+
+  .. grid-item-card:: Install
+
+    * :doc:`Install ComfyUI <install/comfyui-install>`
+
+  .. grid-item-card:: How to
+
+    * :doc:`Download and use models in ComfyUI <how-to/download-models>`
+    * :doc:`Run the Hunyuan3D 2.1 template in ComfyUI <how-to/hunyuan3d-workflow>`
+    * :doc:`Run the Wan 2.2 inference template in ComfyUI <how-to/wan-inference-workflow>`
+
+  .. grid-item-card:: Reference
+
+      * `Overview and reference documentation (upstream) <https://docs.comfy.org/>`__
+
+To contribute to the documentation, refer to
+`Contributing to ComfyUI <https://github.com/AMD-Ecosystem/ComfyUI/blob/master/CONTRIBUTING.md>`__.
+
+You can find licensing information on the :doc:`Licensing <about/license>` page.