- client: Connection management and helpers - tracker: General experiment tracking - inference_tracker: Async metrics for NATS handlers - model_registry: Model registration with KServe metadata - kfp_components: Kubeflow Pipeline components - experiment_comparison: Run comparison tools - cli: Command-line interface
20 lines
423 B
Plaintext
20 lines
423 B
Plaintext
# MLflow Utils Module Requirements
|
|
# Core MLflow
|
|
mlflow>=2.10.0
|
|
|
|
# Database backends
|
|
psycopg2-binary>=2.9.0 # PostgreSQL (CNPG)
|
|
boto3>=1.34.0 # S3-compatible artifact storage (optional)
|
|
|
|
# For async tracking
|
|
aiohttp>=3.9.0
|
|
|
|
# YAML generation for KServe manifests
|
|
PyYAML>=6.0
|
|
|
|
# Already in chat-handler/voice-assistant requirements:
|
|
# httpx (for health checks)
|
|
# Used but typically installed with mlflow:
|
|
# numpy
|
|
# pandas
|