RestClient

Interface for checkpoint management and metadata operations

Training Runs

Get Training Run

get_training_run(run_id)

Retrieve metadata and information about a specific training run.

List Training Runs

list_training_runs(limit=20)

Get paginated list of training runs.

Get by Path

get_training_run_by_tinker_path(path)

Query training run using a MinT path identifier.

Checkpoints

List Checkpoints

list_checkpoints(run_id)

Display training and sampler checkpoints for a run.

Download URL

get_checkpoint_archive_url(checkpoint_id)
get_checkpoint_archive_url_from_tinker_path(path)

Generate signed download URLs for checkpoints.

Example:

future = rc.get_checkpoint_archive_url_from_tinker_path(
    "tinker://<unique_id>/sampler_weights/final"
)
response = future.result()
# response.url is valid until response.expires

Delete Checkpoint

delete_checkpoint(checkpoint_id)

Remove a checkpoint permanently.

Get Weights Info

get_weights_info_by_tinker_path(path)

Extract metadata about saved weights.

Publishing

Publish Checkpoint

publish_checkpoint_from_tinker_path(path)

Make a checkpoint publicly accessible.

Unpublish Checkpoint

unpublish_checkpoint_from_tinker_path(path)

Make a checkpoint private (owner-only access).

Sessions and Samplers

List User Checkpoints

list_user_checkpoints()

Aggregate view of checkpoints across all runs.

Get Session

get_session(session_id)

Retrieve training runs and samplers in a session.

List Sessions

list_sessions(limit=20)

Get paginated list of sessions.

Get Sampler

get_sampler(sampler_id)

Retrieve sampler configuration and metadata.

Async Support

All operations have async variants (e.g., get_training_run_async()).

All methods return ConcurrentFuture or APIFuture - use .result() for blocking or await for async.