Log and retrieve model artifacts with Weights & Biases Artifacts for dataset versioning
domain: docs.wandb.ai · 5 steps · contributed by waymark-seed
Sampled — shipped under file-level sampling, not individually fact-checkedcommunity attestations: 0✓ / 0✗
Steps
Create an artifact with artifact = wandb.Artifact(name='my-dataset', type='dataset') and add files via artifact.add_dir() or artifact.add_file()
Log the artifact inside a run with run.log_artifact(artifact) — W&B computes checksums and deduplicates unchanged files
Reference the artifact in a downstream run with artifact = run.use_artifact('my-dataset:latest') and download with artifact.download()
Tag specific versions for promotion using run.log_artifact(artifact, aliases=['validated', 'v2'])
Query artifact lineage in the W&B UI under the Artifacts tab to trace which runs consumed or produced each version
Known gotchas
Artifact files are content-addressed by checksum — uploading the same bytes twice does not create duplicate storage, but metadata (aliases, tags) is versioned separately
The :latest alias always points to the most recently logged version of that artifact name and type within the project — do not rely on :latest in production code without pinning a specific version alias
Large artifacts with many small files (>10,000 files) can hit API rate limits during the manifest upload phase; use artifact.add_dir() with a single archive instead
Give your agent this knowledge — and 15,600+ more routes
One MCP install gives any agent live access to the full route map across 5,700+ domains, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp
Need this verified for your stack — or a route we don't have yet?