FileNotFoundError: Model stt_fa_fastconformer_hybrid_large was not found.

[csukuangfj]

Describe the bug

FileNotFoundError: Model stt_fa_fastconformer_hybrid_large was not found.

Steps/Code to reproduce bug

import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="stt_fa_fastconformer_hybrid_large")

You can find a colab notebook for reproducing at
https://colab.research.google.com/drive/1x9IeC5bvFa_DTGoH50MVSpVFNQ7FGYoS?usp=sharing

Expected behavior

I am following
https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt_fa_fastconformer_hybrid_large

It should work as expected by following the above link without throwing any errors.

Environment overview (please complete the following information)

Please see
https://colab.research.google.com/drive/1x9IeC5bvFa_DTGoH50MVSpVFNQ7FGYoS?usp=sharing

It should be reproducible under any environment with the master branch.

[csukuangfj]

By the way,

import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="stt_ua_fastconformer_hybrid_large_pc")

from
https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt_ua_fastconformer_hybrid_large_pc

throws the following error:

[NeMo I 2024-05-09 02:22:17 nemo_logging:381] PADDING: 0
[NeMo E 2024-05-09 02:22:17 nemo_logging:405] Model instantiation failed!
    Target class:       nemo.collections.asr.models.hybrid_rnnt_ctc_bpe_models.EncDecHybridRNNTCTCBPEModel
    Error(s):   Error in call to target 'nemo.collections.asr.modules.conformer_encoder.ConformerEncoder':
    TypeError("ConformerEncoder.__init__() got an unexpected keyword argument 'conv_pointwise_type'")
    Traceback (most recent call last):
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py", line 92, in _call_target
        return _target_(*args, **kwargs)
               ^^^^^^^^^^^^^^^^^^^^^^^^^
    TypeError: ConformerEncoder.__init__() got an unexpected keyword argument 'conv_pointwise_type'

    The above exception was the direct cause of the following exception:
    Traceback (most recent call last):
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/nemo/core/classes/common.py", line 502, in from_config_dict
        instance = imported_cls(cfg=config, trainer=trainer)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models.py", line 94, in __init__
        super().__init__(cfg=cfg, trainer=trainer)
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/nemo/collections/asr/models/hybrid_rnnt_ctc_models.py", line 45, in __init__
        super().__init__(cfg=cfg, trainer=trainer)
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/nemo/collections/asr/models/rnnt_models.py", line 66, in __init__
        self.encoder = EncDecRNNTModel.from_config_dict(self.cfg.encoder)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/nemo/core/classes/common.py", line 485, in from_config_dict
        instance = hydra.utils.instantiate(config=config)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py", line 226, in instantiate
        return instantiate_node(
               ^^^^^^^^^^^^^^^^^
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py", line 347, in instantiate_node
        return _call_target(_target_, partial, args, kwargs, full_key)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/mnt/sdb/shared/py311/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py", line 97, in _call_target
        raise InstantiationException(msg) from e
    hydra.errors.InstantiationException: Error in call to target 'nemo.collections.asr.modules.conformer_encoder.ConformerEncoder':
    TypeError("ConformerEncoder.__init__() got an unexpected keyword argument 'conv_pointwise_type'")

[nithinraok]

Hi @csukuangfj , for models published on huggingface you need to prepend nvidia name as well along with model name. so command would look like

import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="nvidia/stt_fa_fastconformer_hybrid_large")

[nithinraok]

I could load stt_ua_fastconformer_hybrid_large_pc fine, could you please try again with: (btw I am on main branch)

asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="nvidia/stt_ua_fastconformer_hybrid_large_pc")

[csukuangfj]

@nithinraok

Thank you for your reply.

nvidia/stt_fa_fastconformer_hybrid_large does work!

I have been following the doc at
https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt_fa_fastconformer_hybrid_large

The doc says stt_fa_fastconformer_hybrid_large, not nvidia/stt_fa_fastconformer_hybrid_large.

I hope that you either fix the doc or fix the NeMo code to support the doc.

[csukuangfj]

for models published on huggingface

How can I figure out whether the model is published on huggingface or not?

I have been using https://catalog.ngc.nvidia.com/models to find models.

[nithinraok]

Looks like there is an issue with NGC model version, and hence I suggested huggingface model its just a copy of it.