check_compatibility¶
yohou_mlflow.check_compatibility(model_uri, extra_trusted_types=None, *, extra_trusted_prefixes=None, extra_version_rules=None, dst_path=None)
¶
Report whether a saved forecaster would load here, without loading it.
Reads only the MLmodel file and the schema stored in the skops file. No object
from the saved forecaster is constructed, so this is safe to run on an untrusted
model, for example in a pre-deploy check.
Parameters ¶
| Name | Type | Description | Default |
|---|---|---|---|
model_uri
|
str
|
Any MLflow model URI. |
required |
extra_trusted_types
|
iterable of str or None
|
Exact type names the eventual load will trust in addition to the policy. |
None
|
extra_trusted_prefixes
|
iterable of str or None
|
Module prefixes the eventual load will trust, each ending in |
None
|
extra_version_rules
|
mapping of str to str or None
|
Version rules the eventual load will apply beyond the built-in ones. |
None
|
dst_path
|
str or None
|
Local directory to download the model into. |
None
|
Returns ¶
| Type | Description |
|---|---|
CompatibilityReport
|
|
Raises ¶
| Type | Description |
|---|---|
ValueError
|
If a trusted prefix is not a module path ending in |
See Also ¶
load_model : Load the forecaster once the report is clean.