feat: update mlflow version to support LLM tracing & update manifests

CHANGELOG.md

@@ -11,6 +11,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### **Changed**
 
+- updated mlflow version to 2.16.0 to support LLM tracing
+- remove CDK overhead from `mlflow-image` module
+- renamed mlflow manifests and updated README.MD
+
 ## v1.5.0
 
 ### **Added**

README.md

@@ -23,15 +23,14 @@ See deployment steps in the [Deployment Guide](DEPLOYMENT.md).
 
 End-to-end example use-cases built using modules in this repository.
 
-| Type                                                                                                      | Description                                                                                                                                                                                                  |
-|-----------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-| [MLOps with Amazon SageMaker](manifests/mlops-sagemaker/)                                                 | Set up environment for MLOps with Amazon SageMaker. Deploy secure Amazon SageMaker Studio Domain, and provisions SageMaker Project Templates using Service Catalog, including model training and deployment. |
-| [Mlflow experiments tracking with Amazon SageMaker](manifests/mlflow-experiments-tracking/)               | An example using Mlflow experiments tracking with Amazon SageMaker. Deploy self-hosted Mlflow instance on AWS Fargate, and Amazon SageMaker Studio Domain environment.                                       |
-| [Managed Workflows with Apache Airflow (MWAA) for Machine Learning Training](manifests/mwaa-ml-training/) | An example orchestrating ML training jobs with Managed Workflows for Apache Airflow (MWAA). Deploys MWAA and an example ML training DAG.                                                                     |
-| [Q&A on PDF documents with RAG](manifests/fmops-qna-rag/)                                                 | Deploy AppSync GraphQL endpoint for Q&A chatbot with RAG based on OpenSearch, and data ingestion infrastructure.                                                                                             |
-| [Ray on Amazon Elastic Kubernetes Service (EKS)](manifests/ray-on-eks/)                                   | Run Ray on AWS EKS. Deploys an AWS EKS cluster, KubeRay Ray Operator, and a Ray Cluster with autoscaling enabled.                                                                                            |
-| [Bedrock Fine-Tuning with Step Functions](manifests/bedrock-finetuning-sfn/)                              | Continuously Fine-tune a Foundation Model with Bedrock Fine-Tuning jobs and AWS Step Functions.                                                                                                              |
-| [MLOps with Step Functions](manifests/mlops-stepfunctions/)                                               | Automate machine learning lifecycle using Amazon SageMaker and AWS Step Functions.                                                                                                                           |
+| Type                                                                                                      | Description                                                                                                                                                                                                                       |
+|-----------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| [MLOps with Amazon SageMaker](manifests/mlops-sagemaker/)                                                 | Set up environment for MLOps with Amazon SageMaker. Deploy secure Amazon SageMaker Studio Domain, and provisions SageMaker Project Templates using Service Catalog, including model training and deployment.                      |
+| [Mlflow tracking server and model registry with Amazon SageMaker](manifests/mlflow-tracking/)             | An example using Mlflow experiments tracking, model registry, and LLM tracing with Amazon SageMaker. Deploy self-hosted Mlflow tracking server and model registry on AWS Fargate, and Amazon SageMaker Studio Domain environment. |
+| [Managed Workflows with Apache Airflow (MWAA) for Machine Learning Training](manifests/mwaa-ml-training/) | An example orchestrating ML training jobs with Managed Workflows for Apache Airflow (MWAA). Deploys MWAA and an example ML training DAG.                                                                                          |
+| [Ray on Amazon Elastic Kubernetes Service (EKS)](manifests/ray-on-eks/)                                   | Run Ray on AWS EKS. Deploys an AWS EKS cluster, KubeRay Ray Operator, and a Ray Cluster with autoscaling enabled.                                                                                                                 |
+| [Bedrock Fine-Tuning with Step Functions](manifests/bedrock-finetuning-sfn/)                              | Continuously Fine-tune a Foundation Model with Bedrock Fine-Tuning jobs and AWS Step Functions.                                                                                                                                   |
+| [MLOps with Step Functions](manifests/mlops-stepfunctions/)                                               | Automate machine learning lifecycle using Amazon SageMaker and AWS Step Functions.                                                                                                                                                |
 
 
 ## Modules

manifests/mlflow-tracking/deployment.yaml

@@ -0,0 +1,23 @@
+name: mlflow-tracking
+toolchainRegion: us-east-1
+forceDependencyRedeploy: true
+groups:
+  - name: networking
+    path: manifests/mlflow-tracking/networking-modules.yaml
+  - name: storage
+    path: manifests/mlflow-tracking/storage-modules.yaml
+  - name: sagemaker-studio
+    path: manifests/mlflow-tracking/sagemaker-studio-modules.yaml
+  - name: images
+    path: manifests/mlflow-tracking/images-modules.yaml
+  - name: mlflow
+    path: manifests/mlflow-tracking/mlflow-modules.yaml
+targetAccountMappings:
+  - alias: primary
+    accountId:
+      valueFrom:
+        envVariable: PRIMARY_ACCOUNT
+    default: true
+    regionMappings:
+      - region: us-east-1
+        default: true

manifests/mlflow-experiments-tracking/images-modules.yaml → manifests/mlflow-tracking/images-modules.yaml

@@ -1,5 +1,5 @@
 name: mlflow-image
-path: git::https://github.com/awslabs/aiops-modules.git//modules/mlflow/mlflow-image?ref=release/1.5.0&depth=1
+path: modules/mlflow/mlflow-image
 targetAccount: primary
 parameters:
   - name: ecr-repository-name

modules/mlflow/mlflow-image/README.md

@@ -2,7 +2,7 @@
 
 ## Description
 
-This module creates an mlflow container image and pushes to the specified Elastic Container Repository.
+This module creates an mlflow tracking server container image and pushes to the specified Elastic Container Repository.
 
 ## Inputs/Outputs
 

modules/mlflow/mlflow-image/deployspec.yaml

@@ -1,30 +1,21 @@
-build_type: BUILD_GENERAL1_SMALL
 publishGenericEnvVariables: true
-
 deploy:
   phases:
-    install:
-      commands:
-        - npm install -g [email protected]
-        - pip install -r requirements.txt
     build:
       commands:
-        - cdk deploy --require-approval never --progress events --app "python app.py" --outputs-file ./cdk-exports.json
-        # Export metadata
-        - seedfarmer metadata convert -f cdk-exports.json || true
-    post_build:
-      commands:
-        - echo "Build successful"
-
+        - aws ecr describe-repositories --repository-names ${SEEDFARMER_PARAMETER_ECR_REPOSITORY_NAME} || aws ecr create-repository --repository-name ${SEEDFARMER_PARAMETER_ECR_REPOSITORY_NAME} --image-scanning-configuration scanOnPush=true
+        - export COMMIT_HASH=$(echo $CODEBUILD_RESOLVED_SOURCE_VERSION | cut -c 1-7)
+        - export IMAGE_TAG=${COMMIT_HASH:=latest}
+        - export REPOSITORY_URI=$AWS_ACCOUNT_ID.dkr.ecr.$AWS_DEFAULT_REGION.amazonaws.com/${SEEDFARMER_PARAMETER_ECR_REPOSITORY_NAME}
+        - aws ecr get-login-password --region $AWS_DEFAULT_REGION | docker login --username AWS --password-stdin $AWS_ACCOUNT_ID.dkr.ecr.$AWS_DEFAULT_REGION.amazonaws.com
+        - echo Building the Docker image...
+        - cd src/ && docker build -t $REPOSITORY_URI:latest .
+        - docker tag $REPOSITORY_URI:latest $REPOSITORY_URI:$IMAGE_TAG
+        - docker push $REPOSITORY_URI:latest && docker push $REPOSITORY_URI:$IMAGE_TAG
+        - seedfarmer metadata add -k ImageUri -v $REPOSITORY_URI:latest
 destroy:
   phases:
-    install:
-      commands:
-        - npm install -g [email protected]
-        - pip install -r requirements.txt
     build:
       commands:
-        - cdk destroy --force --app "python app.py"
-    post_build:
-      commands:
-        - echo "Destroy successful"
+        - aws ecr delete-repository --repository-name ${SEEDFARMER_PARAMETER_ECR_REPOSITORY_NAME} --force
+# build_type: BUILD_GENERAL1_LARGE

modules/mlflow/mlflow-image/modulestack.yaml

@@ -0,0 +1,42 @@
+AWSTemplateFormatVersion: 2010-09-09
+Description: This stack deploys a Module specific IAM permissions
+
+Parameters:
+  # DeploymentName:
+  #   Type: String
+  #   Description: The name of the deployment
+  # ModuleName:
+  #   Type: String
+  #   Description: The name of the Module
+  RoleName:
+    Type: String
+    Description: The name of the IAM Role
+  ECRRepositoryName:
+    Type: String
+    Description: The name of the ECR repository
+
+Resources:
+  Policy:
+    Type: "AWS::IAM::Policy"
+    Properties:
+      PolicyDocument:
+        Statement:
+          - Effect: Allow
+            Action:
+              - "ecr:Describe*"
+              - "ecr:Get*"
+              - "ecr:List*"
+            Resource: "*"
+          - Action:
+              - "ecr:Create*"
+              - "ecr:Delete*"
+              - "ecr:*LayerUpload"
+              - "ecr:UploadLayerPart"
+              - "ecr:Batch*"
+              - "ecr:Put*"
+            Effect: Allow
+            Resource:
+              - !Sub "arn:${AWS::Partition}:ecr:${AWS::Region}:${AWS::AccountId}:repository/${ECRRepositoryName}"
+        Version: 2012-10-17
+      PolicyName: "modulespecific-policy"
+      Roles: [!Ref RoleName]