when i try the tutorial ASR_with_NeMo , AttributeError: 'EncDecCTCModel' object has no attribute 'wer'

[Daniel70-liu]

This error occurs when I perform the inference step. I don't know why. Can anyone kindly teach me?

Bigger batch-size = bigger throughput
params['model']['validation_ds']['batch_size'] = 16

first_asr_model.setup_test_data(test_data_config=params['model']['validation_ds'])
first_asr_model.cuda()
first_asr_model.eval()

wer_nums = []
wer_denoms = []

for test_batch in first_asr_model.test_dataloader():
test_batch = [x.cuda() for x in test_batch]
targets = test_batch[2]
targets_lengths = test_batch[3]
log_probs, encoded_len, greedy_predictions = first_asr_model(
input_signal=test_batch[0], input_signal_length=test_batch[1]
)
first_asr_model.wer.update(predictions=greedy_predictions, predictions_lengths=None, targets=targets, targets_lengths=targets_lengths)
_, wer_num, wer_denom = first_asr_model.wer.compute()
first_asr_model.wer.reset()
wer_nums.append(wer_num.detach().cpu().numpy())
wer_denoms.append(wer_denom.detach().cpu().numpy())
del test_batch, log_probs, targets, targets_lengths, encoded_len, greedy_predictions

print(f"WER = {sum(wer_nums)/sum(wer_denoms)}")

AttributeError Traceback (most recent call last)
Cell In[59], line 29
25 log_probs, encoded_len, greedy_predictions = first_asr_model(
26 input_signal=test_batch[0], input_signal_length=test_batch[1]
27 )
28 # Notice the model has a helper object to compute WER
---> 29 first_asr_model.wer.update(predictions=greedy_predictions, predictions_lengths=None, targets=targets, targets_lengths=targets_lengths)
30 _, wer_num, wer_denom = first_asr_model.wer.compute()
31 first_asr_model.wer.reset()

File ~/anaconda3/envs/nemo/lib/python3.10/site-packages/torch/nn/modules/module.py:1695, in Module.getattr(self, name)
1693 if name in modules:
1694 return modules[name]
-> 1695 raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'")

AttributeError: 'EncDecCTCModel' object has no attribute 'wer'