RuntimeError: input_lengths must be of size batch_size

I am getting the error mentioned in the title while trying to run acoustic model training using wav2vec config Following are the flags used for asr.sh used in run.sh

./asr.sh \
    --stage 10 \
    --stop_stage 10 \
    --feats_normalize null \
    --lang en \
    --audio_format wav \
    --ngpu 4 \
    --nbpe 10000 \
    --max_wav_duration 30 \
    --speed_perturb_factors "0.9 1.0 1.1" \
    --asr_config "${asr_config}" \
    --lm_config "${lm_config}" \
    --inference_config "${inference_config}" \
    --train_set "${train_set}" \
    --valid_set "${valid_set}" \
    --test_sets "${test_sets}" \
    --lm_train_text "data/${train_set}/text" \
    --bpe_train_text "data/${train_set}/text" "$@"

Below is the log of error obtained

Traceback (most recent call last):
  File "/miniconda/envs/espnet/lib/python3.8/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/miniconda/envs/espnet/lib/python3.8/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/espnet/espnet2/tasks/abs_task.py", line 1284, in main_worker
    cls.trainer.run(
  File "/espnet/espnet2/train/trainer.py", line 279, in run
    all_steps_are_invalid = cls.train_one_epoch(
  File "/espnet/espnet2/train/trainer.py", line 499, in train_one_epoch
    retval = model(**batch)
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/parallel/distributed.py", line 705, in forward
    output = self.module(*inputs[0], **kwargs[0])
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/espnet/espnet2/asr/espnet_model.py", line 153, in forward
    loss_ctc, cer_ctc = self._calc_ctc_loss(
  File "/espnet/espnet2/asr/espnet_model.py", line 301, in _calc_ctc_loss
    loss_ctc = self.ctc(encoder_out, encoder_out_lens, ys_pad, ys_pad_lens)
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/espnet/espnet2/asr/ctc.py", line 139, in forward
    loss = self.loss_fn(ys_hat, ys_true, hlens, ys_lens).to(
  File "/espnet/espnet2/asr/ctc.py", line 54, in loss_fn
    loss = self.ctc_loss(th_pred, th_target, th_ilen, th_olen)
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/modules/loss.py", line 1590, in forward
    return F.ctc_loss(log_probs, targets, input_lengths, target_lengths, self.blank, self.reduction,
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/nn/functional.py", line 2307, in ctc_loss
    return torch.ctc_loss(
RuntimeError: input_lengths must be of size batch_size
Traceback (most recent call last):
  File "/miniconda/envs/espnet/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/miniconda/envs/espnet/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/espnet/espnet2/bin/asr_train.py", line 23, in <module>
    main()
  File "/espnet/espnet2/bin/asr_train.py", line 19, in main
    ASRTask.main(cmd=cmd)
  File "/espnet/espnet2/tasks/abs_task.py", line 1044, in main
    while not ProcessContext(processes, error_queues).join():
  File "/miniconda/envs/espnet/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 139, in join
    raise ProcessExitedException(
torch.multiprocessing.spawn.ProcessExitedException: process 1 terminated with exit code 1
# Accounting: time=37 threads=1
# Ended (code 1) at Wed Sep  1 12:48:13 UTC 2021, elapsed time 37 seconds

Can anybody suggest me on this?

[nellorebhanuteja]

Hi I have made the modifications present in the pull request acoustic model training is now working fine!

[simpleoier]

@nellorebhanuteja

Also, are you suggesting all utterances within a batch have same length?

No, I’m suggesting the other way. I suspect the error you posted is because the lengths of sequences in the same batch are all the same. Would you like to try the modifications I made? It’s just two lines.