Reuse a fitted predictor

Fit once, then predict on new rows using the same private workspace identity. Reopening fetches the pinned metadata; it does not upload training data or fit again.

Save the identity#

After a successful fit, retain model.predictor_id_ with your experiment metadata. A predictor ID is an identifier, not a credential or portable executable model.

python

predictor_id = model.predictor_id_
metadata = model.export_metadata()
print(predictor_id, metadata["model_version"])

Metadata export returns authorized JSON-compatible server provenance. It contains no API key, model weights or fitted training context. Do not save prediction payloads, customer rows or credentials into your repository.

Reopen in another session#

Configure your key again for the same workspace. Use the identity you saved and the feature-only test table from Start.

python

from welt import Classifier
from sklearn.metrics import accuracy_score

# Continue with the feature-only Iris test rows from Start.
reopened = Classifier.from_predictor(predictor_id)
repeat = reopened.predict(query)
print("Test accuracy after reopening:", accuracy_score(y_test, repeat))

Use Regressor.from_predictor(...) for a regression predictor. A mismatched task rejects instead of changing the prediction semantics. Reopening keeps model, configuration, version and seed pinned; changing a constructor parameter does not upgrade a fit. A new task/model/configuration needs a new fit.

Inspect your workspace#

python

from welt import Client

with Client() as client:
    predictors = client.predictors()
    metadata = client.predictor(predictor_id)
    events = client.usage()

Predictors and datasets persist until explicit deletion. Idle native cache eviction is execution state, distinct from durable ownership. A reconnect check alone does not prove a worker restart or cold restoration. Usage records show operation, attempt and lifecycle workload metadata; several events may represent one fit. Count distinct operation_id values when counting logical operations. Usage is not a monetary price, credit balance or accuracy report.

See the artifacts and usage notebook for executable checks and waiting and recovery if a fit has not returned an identity.