Dev/mlflow uri handling

README.md

@@ -46,6 +46,44 @@ https://www.kaggle.com/models/emilcode/nightingale
 
 After downloading the folders unzip them.
 
+### How to set up a remote tracking server on your local host system
+https://mlflow.org/docs/3.1.3/ml/tracking/tutorials/remote-server/
+
+https://mlflow.org/docs/3.1.3/ml/tracking/server/
+
+TODO!!!!
+
+### Default tracking server
+If you don't configure/set the MLFLOW_TRACKING_URI to a specific location, MLflow will create a mlruns directory in the CWD or your python run (e.g. next to the notebook if mlfow run has been started for the notebook).
+
+### Use .env file to configure mlflow tracking server
+MLFLOW_TRACKING_URI=<your-tracking-uri> (local file system or url to remote tracking server)
+MLFLOW_EXPERIMENT_ID=<experiment_id>
+
+### Set tracking server during container startup
+Optional: To change the MLFLOW_TRACKING_URI during container startup add something like
+    "containerEnv": {
+        "MLFLOW_TRACKING_URI": "./mlruns"
+    },
+to the .devcontainer.json file.
+
+### Setup mlflow to databricks
+If you have a databricks account and you prefere to track your experiments and register your models to the databricks tracking server and unity catalog your can set up the connection doing the following.
+
+https://docs.databricks.com/aws/en/mlflow3/genai/getting-started/tracing/tracing-ide?language=Use+a+.env+file
+
+Create an .env file in your project root and add:
+'''
+DATABRICKS_TOKEN=<your-databricks-token>
+DATABRICKS_HOST=<host-url-of-your-databricks-cloud> (something like: https://abc-a1234b12-a123.cloud.databricks.com)
+MLFLOW_TRACKING_URI=databricks
+MLFLOW_REGISTRY_URI=databricks-uc (for unity catalog)
+MLFLOW_EXPERIMENT_ID=<experiment_id>
+'''
+
+The .env file will be ignored by git by default, but make sure to NOT commit and push your .env file as it contains credentials not meant to be public. 
+
+
 ### Run mlflow tracking server
 Open the bash/Git bash and change the directory to the unzipped folder containing the model.
 ```bash

notebooks/train_nightingale.ipynb

@@ -285,64 +285,66 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# import mlflow\n",
+    "# from mlflow import MlflowClient\n",
+    "\n",
+    "# TRACKING_URI_LOCAL = \"http://host.docker.internal:5757\"\n",
+    "\n",
+    "# client = MlflowClient(tracking_uri=TRACKING_URI_LOCAL)\n",
+    "\n",
     "import mlflow\n",
-    "from mlflow import MlflowClient\n",
+    "# from mlflow import MlflowClient\n",
     "\n",
-    "TRACKING_URI_LOCAL = \"http://host.docker.internal:5757\"\n",
+    "# At the beginning of your Python script\n",
+    "from dotenv import load_dotenv\n",
     "\n",
-    "client = MlflowClient(tracking_uri=TRACKING_URI_LOCAL)"
+    "# Load environment variables from .env file\n",
+    "load_dotenv()"
    ]
   },
   {
-   "cell_type": "markdown",
-   "id": "348bf482",
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ea50b3d5",
    "metadata": {},
+   "outputs": [],
    "source": [
-    "### Create experiment \n",
-    "RUN THE FOLLOWING CODE BLOCK ONLY ONCE FOR INITIAL EXPERIMENT SETUP!!"
+    "import os\n",
+    "print(\"Env:\", os.getenv(\"MLFLOW_TRACKING_URI\"))\n",
+    "print(\"From MLflow:\", mlflow.get_tracking_uri())"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "1dcec8ff",
+   "cell_type": "markdown",
+   "id": "348bf482",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "experiment_description = (\n",
-    "    \"Nightingale is a bird call classification project.\"\n",
-    ")\n",
-    "\n",
-    "experiment_tags = {\n",
-    "    \"project_name\": \"nightingale\",\n",
-    "    \"mlflow.note.content\": experiment_description,\n",
-    "}\n",
-    "\n",
-    "# only run following command once to create the experiment after the server has been started for the first time\n",
-    "client.create_experiment(name=\"Nightingale Bird Call Classification\", tags=experiment_tags)"
+    "### Create experiment \n",
+    "RUN THE FOLLOWING CODE BLOCK ONLY ONCE FOR INITIAL EXPERIMENT SETUP!!"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "a5010b9c",
+   "id": "1dcec8ff",
    "metadata": {},
    "outputs": [],
    "source": [
-    "\n",
-    "\n",
-    "# Use the fluent API to set the tracking uri and the active experiment\n",
-    "mlflow.set_tracking_uri(TRACKING_URI_LOCAL)\n",
-    "\n",
-    "# Sets the current active experiment to the \"Nightingale_Bird_Call_Classification\" experiment and returns the Experiment metadata\n",
-    "nightingale_experiment = mlflow.set_experiment(\"Nightingale Bird Call Classification\")\n",
-    "\n",
-    "# Define a run name for this iteration of training.\n",
-    "# If this is not set, a unique name will be auto-generated for your run.\n",
-    "# run_name = \"nightingale_classifier_test\"\n",
-    "\n",
-    "# Define an artifact path that the model will be saved to.\n",
-    "# artifact_path = \"classifier_nightingale\""
+    "# experiment_description = (\n",
+    "#     \"Nightingale is a bird call classification project.\"\n",
+    "# )\n",
+    "\n",
+    "# experiment_tags = {\n",
+    "#     \"project_name\": \"nightingale\",\n",
+    "#     \"mlflow.note.content\": experiment_description,\n",
+    "# }\n",
+    "\n",
+    "# # only run following command once to create the experiment after the server has been started for the first time\n",
+    "# # client.create_experiment(name=\"Nightingale Bird Call Classification\", tags=experiment_tags)\n",
+    "# mlflow.set_experiment(\n",
+    "#     experiment_name=\"/Workspace/Users/ephraim.eckl@posteo.de/nightingale\",\n",
+    "#     experiment_id=\"2165278269360514\"\n",
+    "# )"
    ]
   },
   {
@@ -427,7 +429,7 @@
     "\n",
     "    print(\"Shape of input_example:\", sample_input.shape)\n",
     "    # Log an instance of the trained model for later use\n",
-    "    model_info = mlflow.keras.log_model(model=bird_class_model, name = \"Bird-Call-Classifier-Head\", signature=signature, pip_requirements=['keras==3.10.0'], registered_model_name=\"Reg-Bird-Call-Classifier-Head\")\n",
+    "    model_info = mlflow.keras.log_model(model=bird_class_model, name = \"Bird-Call-Classifier-Head\", signature=signature, pip_requirements=['keras==3.10.0'], registered_model_name=\"nightingale-dev.default.Reg-Bird-Call-Classifier-Head\")\n",
     "#     # mlflow.sklearn.log_model(sk_model=rf, input_example=X_val, name=artifact_path)\n",
     "    "
    ]
@@ -462,7 +464,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "bird_class_model.save('bird_classifier_head.keras')"
+    "# bird_class_model.save('bird_classifier_head.keras')"
    ]
   }
  ],