Problem using single_file_inference.sh

[vpurandara]

Installed vakyansh-wav2vec2-experimentation as mentioned in ReadMe.md

Also downloaded the Vakyansh Open Source Models.

No GPUs available.. Running in CPU.

scripts/inference/bash single_file_inference.sh

while executing..

1. if Single Model for Inference is given for custom_model_path then following error occured....

(vwav2vec2) purandara@purandara:/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/scripts/inference$ bash single_file_inference.sh
Traceback (most recent call last):
File "../../utils/inference/single_file_inference.py", line 380, in
result = parse_transcription(args_local.model, args_local.dict, args_local.wav, args_local.cuda, args_local.decoder, args_local.lexicon, args_local.lm_path, args_local.half)
File "../../utils/inference/single_file_inference.py", line 363, in parse_transcription
result = get_results(wav_path=wav_path, dict_path=dict_path, generator=generator, use_cuda=cuda, model=model, half=half)
File "../../utils/inference/single_file_inference.py", line 315, in get_results
hypo = generator.generate(model, sample, prefix_tokens=None)
File "../../utils/inference/single_file_inference.py", line 113, in generate
emissions = self.get_emissions(models, encoder_input)
File "../../utils/inference/single_file_inference.py", line 119, in get_emissions
encoder_out = model(**encoder_input)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "../../utils/inference/single_file_inference.py", line 70, in forward
x = self.w2v_encoder(**kwargs)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/fairseq/fairseq/models/wav2vec/wav2vec2_asr.py", line 484, in forward
res = self.w2v_model.extract_features(**w2v_args)
File "/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/fairseq/fairseq/models/wav2vec/wav2vec2.py", line 778, in extract_features
res = self.forward(
File "/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/fairseq/fairseq/models/wav2vec/wav2vec2.py", line 599, in forward
features = self.feature_extractor(source)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/fairseq/fairseq/models/wav2vec/wav2vec2.py", line 895, in forward
x = conv(x)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/container.py", line 139, in forward
input = module(input)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/activation.py", line 681, in forward
return F.gelu(input, approximate=self.approximate)
File "/home/purandara/anaconda3/envs/vwav2vec2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1207, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'GELU' object has no attribute 'approximate'

2. if Finetuned Model is given for custom_model_path then following error occured....

(vwav2vec2) purandara@purandara:/media/purandara/harddisk/VWAV2VEC2/vakyansh-wav2vec2-experimentation/scripts/inference$ bash single_file_inference.sh
Traceback (most recent call last):
File "../../utils/inference/single_file_inference.py", line 380, in
result = parse_transcription(args_local.model, args_local.dict, args_local.wav, args_local.cuda, args_local.decoder, args_local.lexicon, args_local.lm_path, args_local.half)
File "../../utils/inference/single_file_inference.py", line 363, in parse_transcription
result = get_results(wav_path=wav_path, dict_path=dict_path, generator=generator, use_cuda=cuda, model=model, half=half)
File "../../utils/inference/single_file_inference.py", line 301, in get_results
model.eval()
AttributeError: 'dict' object has no attribute 'eval'

These errors are happened for both decoders i.e; viterbi and kenlm

I dont know where I went wrong. Can anyone help me with it.

Note: I installed flashlight using git clone https://github.com/flashlight/flashlight.git --branch v0.3.2 for this Reason