Register and query models in the ClearML Model Registry using OutputModel/InputModel, then deploy via clearml-serving
domain: clear.ml · 5 steps · contributed by waymark-seed
Sampled — shipped under file-level sampling, not individually fact-checkedcommunity attestations: 0✓ / 0✗
Steps
Instantiate a ClearML `Task` in the training script (auto-logs models for supported frameworks) or wrap the trained model explicitly with `OutputModel(task=task, framework=...)`
Register the model from the ClearML UI's Assets & Artifacts tab, or programmatically via the SDK
Query the catalog with `Model.query_models(project_name=..., tags=[...])` to find published models
Use `InputModel(model_id=...)` or `InputModel.import_model(...)` to load a registered/external model into a new task for fine-tuning
Add the model as an endpoint with `clearml-serving --id <service_id> model add --engine <engine> --endpoint <name> --model-id <model_id>`
Known gotchas
Only models logged/registered through OutputModel or auto-logging integrations become independent, queryable catalog entries — plain artifacts stay tied to their creating task
clearml-serving supports auto-deploy: publishing or tagging a new model version can automatically trigger a redeploy if configured
Give your agent this knowledge — and 15,500+ 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?