Looking for usage documentation? Check out Classification and Regression.
FittedModelNotFoundError
View source predict when model_id_ — set by fit(), restored by
load_model() or assigned directly — is unknown to the server: the fitted
model was deleted (for instance through UserDataClient), or it belongs to
a different account. Call fit() again to create a new one.
InteractiveLoginError
View source foundry.FoundryEndpointError
View source httpx.HTTPStatusError, so callers that already catch that
keep working. The endpoint reports failures as a JSON body carrying
error_code and trace_id; both are surfaced on the exception (and in
its message) because the trace_id is what the endpoint provider needs
in order to look the failure up server-side.
FitFailedError
View source fit() and wait_for_fit() when they poll the server for the
outcome of a fit, as fit() does in ASYNC mode. The message holds the
server’s reason.
PredictFailedError
View source predict(), submit_predict() and wait_for_predict(),
whether the server reports the failure in answer to the prediction request
itself or to a poll for the prediction. The message holds the server’s
reason and its error code, and fitted_test_set_id the id of the
prediction.
Initialize the error.
RefitRequiredError
View source fit() again.
Raised by predict(), submit_predict() and wait_for_predict()
when the server cannot predict with the fit model_id_ refers to: an
earlier version of the service created it, or its cached state is no longer
valid. The server’s error code says which: NEEDS_REFITTING or
KV_CACHE_INVALID. The message holds the server’s reason.
Initialize the error.