You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In test_push_to_hub_model_kwargs and test_save_pretrained_kwargs_load_fails_without_traces, it seems that the tests are not working as expected. They should apparently check that loading from pretrained weights raises an exception if custom layers are used.
Tests implementation are:
(...)
from_pretrained_keras(f"{WORKING_REPO_DIR}/{REPO_NAME}")
self.assertRaises(ValueError, msg="Exception encountered when calling layer*")
instead of something like:
(...)
withself.assertRaises(ValueError, msg="Exception encountered when calling layer*"):
from_pretrained_keras(f"{WORKING_REPO_DIR}/{REPO_NAME}")
In the end, the from_pretrained_keras call works without exception. @merveenoyan@osanseviero I am not very familiar with the "custom layer problem" in Keras but I saw the tests have been introduced by this PR. Could you have a look and see if there is an actual problem with the layers ? Thanks in advance !
In
test_push_to_hub_model_kwargs
andtest_save_pretrained_kwargs_load_fails_without_traces
, it seems that the tests are not working as expected. They should apparently check that loading from pretrained weights raises an exception if custom layers are used.Tests implementation are:
instead of something like:
In the end, the
from_pretrained_keras
call works without exception. @merveenoyan @osanseviero I am not very familiar with the "custom layer problem" in Keras but I saw the tests have been introduced by this PR. Could you have a look and see if there is an actual problem with the layers ? Thanks in advance !You can find test implementations here and here.
The text was updated successfully, but these errors were encountered: