ML experiment tracking, dataset versioning, and model registry.
## Weights & Biases MCP Server: ML Experiment Tracking The **Weights & Biases (W&B) MCP Server** integrates the industry-leading ML experiment tracking platform into Google Antigravity. This tool enables comprehensive logging, visualization, and collaboration for machine learning experiments at any scale. ### Why W&B MCP? W&B is the gold standard for ML tracking: - **Experiment Tracking**: Log everything automatically - **Visualizations**: Beautiful interactive dashboards - **Collaboration**: Share experiments with team - **Artifacts**: Version datasets and models - **Sweeps**: Automated hyperparameter optimization ### Key Features #### 1. Experiment Logging ```python import wandb # Initialize run wandb.init(project="my-project", config={ "learning_rate": 0.001, "epochs": 100, "batch_size": 32 }) # Training loop for epoch in range(100): train_loss = train_epoch(model) val_loss = validate(model) # Log metrics wandb.log({ "epoch": epoch, "train_loss": train_loss, "val_loss": val_loss, "learning_rate": scheduler.get_lr() }) wandb.finish() ``` #### 2. Hyperparameter Sweeps ```python # Define sweep configuration sweep_config = { "method": "bayes", "metric": {"name": "val_loss", "goal": "minimize"}, "parameters": { "learning_rate": {"min": 0.0001, "max": 0.1}, "batch_size": {"values": [16, 32, 64]}, "hidden_size": {"values": [128, 256, 512]} } } sweep_id = wandb.sweep(sweep_config, project="my-sweeps") def train(): wandb.init() config = wandb.config model = create_model(config.hidden_size) # Train with config params... wandb.agent(sweep_id, train, count=50) ``` #### 3. Artifact Management ```python # Log dataset as artifact artifact = wandb.Artifact("training-data", type="dataset") artifact.add_dir("./data/") wandb.log_artifact(artifact) # Log model as artifact model_artifact = wandb.Artifact("trained-model", type="model") model_artifact.add_file("model.pt") wandb.log_artifact(model_artifact) # Download artifact in another run artifact = wandb.use_artifact("training-data:latest") artifact_dir = artifact.download() ``` ### Configuration ```json { "mcpServers": { "wandb": { "command": "npx", "args": ["-y", "@anthropic/mcp-wandb"], "env": { "WANDB_API_KEY": "your-api-key", "WANDB_PROJECT": "your-project" } } } } ``` ### Use Cases **Experiment Comparison**: Compare hundreds of experiments to identify best configurations. **Model Registry**: Track model lineage from training data to production deployment. **Team Collaboration**: Share dashboards and reports with stakeholders. The W&B MCP Server brings world-class ML experiment tracking to Antigravity.
{
"mcpServers": {
"wandb": {}
}
}