Derk Mus

And for my understanding, is it expected that the forward method of the `VMambaForImageClassification` returns a ImageClassifierOutputWithNoAttention object? And the VMamba model should return a `BaseModelOutput`?

Two tests that are still failing are the `VMambaModelTest::test_torch_fx` and `VMambaModelTest::test_torch_fx_output_loss` because AssertionError: Couldn't trace module: Model VMambaModel is not supported yet, supported models: AlbertForMaskedLM, AlbertForMultipleChoice, AlbertForPreTraining, AlbertForQuestionAnswering, AlbertForSequenceClassificat... I...

Also the `test_initialization` is failing. This check that fails is if param.requires_grad: self.assertIn( ((param.data.mean() * 1e9).round() / 1e9).item(), [0.0, 1.0], msg=f"Parameter {name} of model {model_class} seems not properly initialized", )...

All test are passing now locally. (I still use a local checkpoint because the vmamba checkpoint is not uploaded to the hub yet). So I think this PR is ready...

For this error: > Checking all objects are properly documented. > Traceback (most recent call last): > File "/home/derk/transformers/utils/check_repo.py", line 1181, in > check_repo_quality() > File "/home/derk/transformers/utils/check_repo.py", line 1165, in...

Thanks. Now I encounter this error when running python utils/check_inits.py: Traceback (most recent call last): File "/home/derk/transformers/utils/check_inits.py", line 370, in check_all_inits() File "/home/derk/transformers/utils/check_inits.py", line 298, in check_all_inits raise ValueError("\n\n".join(failures)) ValueError:...

Yes, stil on the radar. Will try to update it soon Op za 23 mrt 2024 01:48 schreef Alessio Placitelli ***@***.*** >: > This PR is very exciting! Glad to...

> Hi @dmus - any updates on this PR? I now encounter the following error while running the tests. > Traceback (most recent call last): > File "/home/derk/transformers/utils/check_inits.py", line 370,...