google-cloud-mldiagnostics

import os import logging import google.cloud.logging from google_cloud_mldiagnostics import machinelearning_run from google_cloud_mldiagnostics import metrics from google_cloud_mldiagnostics import xprof from google_cloud_mldiagnostics.proto.diagnostics import MetricType # Set up Cloud Logging (recommended) logging_client = google.cloud.logging.Client() logging_client.setup_logging() logging.info("Cloud Logging is set up.") PROJECT_ID = os.environ.get('GCP_PROJECT_ID', 'your-gcp-project-id') # Ensure GOOGLE_APPLICATION_CREDENTIALS is set or authenticated via gcloud CLI def run_ml_diagnostics_example(): print(f"Using GCP Project ID: {PROJECT_ID}") # 1. Create a machine learning run # The SDK automatically generates a unique run_id if not provided. run_name = machinelearning_run.create_run( project_id=PROJECT_ID, experiment_name="my-first-experiment", display_name="my-training-run" ) print(f"Created ML Run: {run_name}") # 2. Record metrics metrics.record(MetricType.LOSS, 0.5, step=1, run_name=run_name) metrics.record(MetricType.ACCURACY, 0.8, step=1, run_name=run_name) print("Recorded initial metrics.") metrics.record(MetricType.LOSS, 0.2, step=10, run_name=run_name) metrics.record(MetricType.ACCURACY, 0.95, step=10, run_name=run_name) print("Recorded updated metrics.") # 3. Write configurations (example) machinelearning_run.write_config(run_name, {"learning_rate": 0.01, "batch_size": 32}) print("Wrote run configurations.") # Example of capturing a profile (requires XProf server running in your workload) # For on-demand capture, ensure xprof.start_server() is called in your ML workload. # xprof.capture_profile(run_name, 'gs://your-bucket/profiles', duration_ms=10000) # print("Attempted to capture profile.") print("ML Diagnostics example completed. Check Google Cloud Console for 'my-training-run'.") if __name__ == '__main__': run_ml_diagnostics_example()