docs: address PR feedback - deduplicate MLflow content and update screenshots

briangallagher · cursoragent · briangallagher · commit 5b069d1e35a3 · 2026-06-09T08:04:39.000+01:00
- Extract shared MLflow documentation into examples/fine-tuning/mlflow.md
- Replace duplicated content in lora/osft/sft READMEs with link to shared doc
- Update screenshots: remove email from top-right, use generic "fine-tuning"
  experiment name instead of method-specific names

Co-authored-by: Cursor &lt;cursoragent@cursor.com&gt;
diff --git a/examples/fine-tuning/images/mlflow-experiments.png b/examples/fine-tuning/images/mlflow-experiments.png
diff --git a/examples/fine-tuning/images/mlflow-run-metrics.png b/examples/fine-tuning/images/mlflow-run-metrics.png
diff --git a/examples/fine-tuning/lora/README.md b/examples/fine-tuning/lora/README.md
@@ -175,66 +175,4 @@ You can now proceed with the instructions from the notebook. Enjoy!
 
 ## MLflow Integration (Optional)
 
-Training Hub supports [MLflow](https://mlflow.org/) for experiment tracking. When MLflow is enabled on your RHOAI cluster, training metrics (loss, learning rate, etc.) are automatically logged to MLflow experiments — no additional code changes required beyond setting the experiment name.
-
-> [!NOTE]
-> MLflow integration is available for **interactive (single node)** notebooks only. Distributed training jobs do not currently support MLflow tracking.
-
-### Enabling MLflow
-
-The interactive notebook already includes a cell that sets the MLflow experiment name:
-
-```python
-os.environ["MLFLOW_EXPERIMENT_NAME"] = "lora-training"
-```
-
-For this to work, MLflow must be enabled as a component in your RHOAI installation. If MLflow is not enabled, the environment variable is simply ignored and training proceeds normally.
-
-**To enable MLflow on your cluster:**
-
-1. Enable the MLflow Operator component in your `DataScienceCluster` CR:
-
-   ```bash
-   oc patch datasciencecluster default-dsc \
-     --type=merge \
-     -p '{"spec":{"components":{"mlflowoperator":{"managementState":"Managed"}}}}'
-   ```
-
-2. Create an `MLflow` CR to deploy the tracking server (example using SQLite and a PV for storage):
-
-   ```bash
-   oc apply -f - <<EOF
-   apiVersion: mlflow.opendatahub.io/v1
-   kind: MLflow
-   metadata:
-     name: mlflow
-   spec:
-     backendStoreUri: "sqlite:////mlflow/mlflow.db"
-     defaultArtifactRoot: "file:///mlflow/artifacts"
-     serveArtifacts: true
-     storage:
-       accessModes:
-         - ReadWriteOnce
-       resources:
-         requests:
-           storage: 10Gi
-   EOF
-   ```
-
-For full details, see the [Configuring MLflow in OpenShift AI](https://access.redhat.com/articles/7136121) Knowledgebase article.
-
-### Viewing MLflow Experiments
-
-Once training completes with MLflow enabled, you can browse your experiment runs:
-
-1. In the OpenShift AI dashboard, navigate to **Develop & train → Experiments** from the left sidebar menu.
-2. Select the experiment name (e.g., `lora-training`) to view all runs.
-3. Each run contains logged metrics (training loss, learning rate), parameters, and artifacts.
-
-You can also launch the full MLflow UI by clicking the **"Launch MLflow"** link in the top right of the Experiments page:
-
-![](../images/mlflow-experiments.png)
-
-Each run logs metrics including training loss, learning rate, samples per second, and more:
-
-![](../images/mlflow-run-metrics.png)
+The interactive notebook supports optional MLflow experiment tracking. See the [MLflow Integration guide](../mlflow.md) for setup instructions and details.
diff --git a/examples/fine-tuning/mlflow.md b/examples/fine-tuning/mlflow.md
@@ -0,0 +1,65 @@
+# MLflow Integration (Optional)
+
+Training Hub supports [MLflow](https://mlflow.org/) for experiment tracking. When MLflow is enabled on your RHOAI cluster, training metrics (loss, learning rate, etc.) are automatically logged to MLflow experiments — no additional code changes required beyond setting the experiment name.
+
+> [!NOTE]
+> MLflow integration is available for **interactive (single node)** notebooks only. Distributed training jobs do not currently support MLflow tracking.
+
+## Enabling MLflow
+
+Each interactive notebook already includes a cell that sets the MLflow experiment name:
+
+```python
+os.environ["MLFLOW_EXPERIMENT_NAME"] = "<your-experiment-name>"
+```
+
+For this to work, MLflow must be enabled as a component in your RHOAI installation. If MLflow is not enabled, the environment variable is simply ignored and training proceeds normally.
+
+**To enable MLflow on your cluster:**
+
+1. Enable the MLflow Operator component in your `DataScienceCluster` CR:
+
+   ```bash
+   oc patch datasciencecluster default-dsc \
+     --type=merge \
+     -p '{"spec":{"components":{"mlflowoperator":{"managementState":"Managed"}}}}'
+   ```
+
+2. Create an `MLflow` CR to deploy the tracking server (example using SQLite and a PV for storage):
+
+   ```bash
+   oc apply -f - <<EOF
+   apiVersion: mlflow.opendatahub.io/v1
+   kind: MLflow
+   metadata:
+     name: mlflow
+   spec:
+     backendStoreUri: "sqlite:////mlflow/mlflow.db"
+     defaultArtifactRoot: "file:///mlflow/artifacts"
+     serveArtifacts: true
+     storage:
+       accessModes:
+         - ReadWriteOnce
+       resources:
+         requests:
+           storage: 10Gi
+   EOF
+   ```
+
+For full details, see the [Configuring MLflow in OpenShift AI](https://access.redhat.com/articles/7136121) Knowledgebase article.
+
+## Viewing MLflow Experiments
+
+Once training completes with MLflow enabled, you can browse your experiment runs:
+
+1. In the OpenShift AI dashboard, navigate to **Develop & train → Experiments** from the left sidebar menu.
+2. Select the experiment name to view all runs.
+3. Each run contains logged metrics (training loss, learning rate), parameters, and artifacts.
+
+You can also launch the full MLflow UI by clicking the **"Launch MLflow"** link in the top right of the Experiments page:
+
+![MLflow experiments page](./images/mlflow-experiments.png)
+
+Each run logs metrics including training loss, learning rate, samples per second, and more:
+
+![MLflow run metrics](./images/mlflow-run-metrics.png)
diff --git a/examples/fine-tuning/osft/README.md b/examples/fine-tuning/osft/README.md
@@ -210,66 +210,4 @@ You can now proceed with the instructions from the notebook. Enjoy!
 
 ## MLflow Integration (Optional)
 
-Training Hub supports [MLflow](https://mlflow.org/) for experiment tracking. When MLflow is enabled on your RHOAI cluster, training metrics (loss, learning rate, etc.) are automatically logged to MLflow experiments — no additional code changes required beyond setting the experiment name.
-
-> [!NOTE]
-> MLflow integration is available for **interactive (single node)** notebooks only. Distributed training jobs do not currently support MLflow tracking.
-
-### Enabling MLflow
-
-The interactive notebook already includes a cell that sets the MLflow experiment name:
-
-```python
-os.environ["MLFLOW_EXPERIMENT_NAME"] = "osft-training"
-```
-
-For this to work, MLflow must be enabled as a component in your RHOAI installation. If MLflow is not enabled, the environment variable is simply ignored and training proceeds normally.
-
-**To enable MLflow on your cluster:**
-
-1. Enable the MLflow Operator component in your `DataScienceCluster` CR:
-
-   ```bash
-   oc patch datasciencecluster default-dsc \
-     --type=merge \
-     -p '{"spec":{"components":{"mlflowoperator":{"managementState":"Managed"}}}}'
-   ```
-
-2. Create an `MLflow` CR to deploy the tracking server (example using SQLite and a PV for storage):
-
-   ```bash
-   oc apply -f - <<EOF
-   apiVersion: mlflow.opendatahub.io/v1
-   kind: MLflow
-   metadata:
-     name: mlflow
-   spec:
-     backendStoreUri: "sqlite:////mlflow/mlflow.db"
-     defaultArtifactRoot: "file:///mlflow/artifacts"
-     serveArtifacts: true
-     storage:
-       accessModes:
-         - ReadWriteOnce
-       resources:
-         requests:
-           storage: 10Gi
-   EOF
-   ```
-
-For full details, see the [Configuring MLflow in OpenShift AI](https://access.redhat.com/articles/7136121) Knowledgebase article.
-
-### Viewing MLflow Experiments
-
-Once training completes with MLflow enabled, you can browse your experiment runs:
-
-1. In the OpenShift AI dashboard, navigate to **Develop & train → Experiments** from the left sidebar menu.
-2. Select the experiment name (e.g., `osft-training`) to view all runs.
-3. Each run contains logged metrics (training loss, learning rate), parameters, and artifacts.
-
-You can also launch the full MLflow UI by clicking the **"Launch MLflow"** link in the top right of the Experiments page:
-
-![](../images/mlflow-experiments.png)
-
-Each run logs metrics including training loss, learning rate, samples per second, and more:
-
-![](../images/mlflow-run-metrics.png)
+The interactive notebook supports optional MLflow experiment tracking. See the [MLflow Integration guide](../mlflow.md) for setup instructions and details.
diff --git a/examples/fine-tuning/sft/README.md b/examples/fine-tuning/sft/README.md
@@ -157,66 +157,4 @@ You can now proceed with the instructions from the notebook. Enjoy!
 
 ## MLflow Integration (Optional)
 
-Training Hub supports [MLflow](https://mlflow.org/) for experiment tracking. When MLflow is enabled on your RHOAI cluster, training metrics (loss, learning rate, etc.) are automatically logged to MLflow experiments — no additional code changes required beyond setting the experiment name.
-
-> [!NOTE]
-> MLflow integration is available for **interactive (single node)** notebooks only. Distributed training jobs do not currently support MLflow tracking.
-
-### Enabling MLflow
-
-The interactive notebook already includes a cell that sets the MLflow experiment name:
-
-```python
-os.environ["MLFLOW_EXPERIMENT_NAME"] = "sft-training"
-```
-
-For this to work, MLflow must be enabled as a component in your RHOAI installation. If MLflow is not enabled, the environment variable is simply ignored and training proceeds normally.
-
-**To enable MLflow on your cluster:**
-
-1. Enable the MLflow Operator component in your `DataScienceCluster` CR:
-
-   ```bash
-   oc patch datasciencecluster default-dsc \
-     --type=merge \
-     -p '{"spec":{"components":{"mlflowoperator":{"managementState":"Managed"}}}}'
-   ```
-
-2. Create an `MLflow` CR to deploy the tracking server (example using SQLite and a PV for storage):
-
-   ```bash
-   oc apply -f - <<EOF
-   apiVersion: mlflow.opendatahub.io/v1
-   kind: MLflow
-   metadata:
-     name: mlflow
-   spec:
-     backendStoreUri: "sqlite:////mlflow/mlflow.db"
-     defaultArtifactRoot: "file:///mlflow/artifacts"
-     serveArtifacts: true
-     storage:
-       accessModes:
-         - ReadWriteOnce
-       resources:
-         requests:
-           storage: 10Gi
-   EOF
-   ```
-
-For full details, see the [Configuring MLflow in OpenShift AI](https://access.redhat.com/articles/7136121) Knowledgebase article.
-
-### Viewing MLflow Experiments
-
-Once training completes with MLflow enabled, you can browse your experiment runs:
-
-1. In the OpenShift AI dashboard, navigate to **Develop & train → Experiments** from the left sidebar menu.
-2. Select the experiment name (e.g., `sft-training`) to view all runs.
-3. Each run contains logged metrics (training loss, learning rate), parameters, and artifacts.
-
-You can also launch the full MLflow UI by clicking the **"Launch MLflow"** link in the top right of the Experiments page:
-
-![](../images/mlflow-experiments.png)
-
-Each run logs metrics including training loss, learning rate, samples per second, and more:
-
-![](../images/mlflow-run-metrics.png)
+The interactive notebook supports optional MLflow experiment tracking. See the [MLflow Integration guide](../mlflow.md) for setup instructions and details.
