Convasr_to_enemo.py is throwing error

bjoish · April 30, 2021, 1:17pm

I am trying to convert Nemo to jarvis by following up this docs
https://docs.nvidia.com/deeplearning/jarvis/user-guide/docs/model-overview.html?highlight=enemo#nemo-to-jarvis

When invoking the convasr_to_enemo.py, I am getting error

./model_config.yaml not found

But I am able to see model_config.yaml inside the QuartzNet15x5.nemo (I renamed QuartzNet15x5.tar)

cmd:
python convasr_to_enemo.py --nemo_file=./QuartzNet15x5.nemo --onnx_file=./quartz.onnx --enemo_file=./quartznet_asr.enemo

error:

Traceback (most recent call last):
File “jarvis.py”, line 74, in
main(
File “jarvis.py”, line 64, in main
archive.extract(‘./model_config.yaml’)
File “/usr/lib/python3.8/tarfile.py”, line 2058, in extract
tarinfo = self.getmember(member)
File “/usr/lib/python3.8/tarfile.py”, line 1780, in getmember
raise KeyError(“filename %r not found” % name)
KeyError: “filename ‘./model_config.yaml’ not found”

SunilJB · May 3, 2021, 6:29pm

Hi @bjoish,

Can you try using original model naming convention "QuartzNet15x5Base-En"

It seems pretrained model is downloaded as per the mapping defined in the below link:

github.com

NVIDIA/NeMo/blob/855ce265b80c0dc40f4f06ece76d2c9d6ca1be8d/nemo/collections/asr/models/ctc_models.py#L52


      
          @classmethod
          def list_available_models(cls) -> Optional[PretrainedModelInfo]:
              """
              This method returns a list of pre-trained model which can be instantiated directly from NVIDIA's NGC cloud.
          
              Returns:
                  List of available pre-trained models.
              """
              results = []
          
              model = PretrainedModelInfo(
                  pretrained_model_name="QuartzNet15x5Base-En",
                  description="QuartzNet15x5 model trained on six datasets: LibriSpeech, Mozilla Common Voice (validated clips from en_1488h_2019-12-10), WSJ, Fisher, Switchboard, and NSC Singapore English. It was trained with Apex/Amp optimization level O1 for 600 epochs. The model achieves a WER of 3.79% on LibriSpeech dev-clean, and a WER of 10.05% on dev-other. Please visit https://ngc.nvidia.com/catalog/models/nvidia:nemospeechmodels for further details.",
                  location="https://api.ngc.nvidia.com/v2/models/nvidia/nemospeechmodels/versions/1.0.0a5/files/QuartzNet15x5Base-En.nemo",
              )
              results.append(model)
          
              model = PretrainedModelInfo(
                  pretrained_model_name="stt_en_quartznet15x5",
                  description="For details about this model, please visit https://ngc.nvidia.com/catalog/models/nvidia:nemo:stt_en_quartznet15x5",
                  location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/stt_en_quartznet15x5/versions/1.0.0rc1/files/stt_en_quartznet15x5.nemo",

Thanks