[Bug]: RuntimeError: The size of tensor a (140590) must match the size of tensor b (2500) at non-singleton dimension 1 #2071

from speechbrain.pretrained import EncoderDecoderASR

asr_model = EncoderDecoderASR.from_hparams(source="speechbrain/asr-transformer-transformerlm-librispeech", savedir="pretrained_models/asr-transformer-transformerlm-librispeech")
trans = asr_model.transcribe_file("my/local/path/file.wav")

RuntimeError                              Traceback (most recent call last)
Cell In[10], line 3
      1 with torch.no_grad():
      2     wav_lens = rel_length.to(asr_model.device)
----> 3     encoder_out = asr_model.encode_batch(batch, rel_length)

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\speechbrain\pretrained\interfaces.py:604](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/speechbrain/pretrained/interfaces.py:604), in EncoderDecoderASR.encode_batch(self, wavs, wav_lens)
    602 wavs = wavs.float()
    603 wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
--> 604 encoder_out = self.mods.encoder(wavs, wav_lens)
    605 return encoder_out

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\torch\nn\modules\module.py:1501](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/torch/nn/modules/module.py:1501), in Module._call_impl(self, *args, **kwargs)
   1496 # If we don't have any hooks, we want to skip the rest of the logic in
   1497 # this function, and just call forward.
   1498 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1499         or _global_backward_pre_hooks or _global_backward_hooks
   1500         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501     return forward_call(*args, **kwargs)
   1502 # Do not call functions when jit is used
   1503 full_backward_hooks, non_full_backward_hooks = [], []

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\speechbrain\nnet\containers.py:191](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/speechbrain/nnet/containers.py:191), in LengthsCapableSequential.forward(self, x, lengths)
    189     x = layer(x, lengths=lengths)
    190 else:
--> 191     x = layer(x)
    192 if isinstance(x, tuple):
    193     x = x[0]

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\torch\nn\modules\module.py:1501](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/torch/nn/modules/module.py:1501), in Module._call_impl(self, *args, **kwargs)
   1496 # If we don't have any hooks, we want to skip the rest of the logic in
   1497 # this function, and just call forward.
   1498 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1499         or _global_backward_pre_hooks or _global_backward_hooks
   1500         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501     return forward_call(*args, **kwargs)
   1502 # Do not call functions when jit is used
   1503 full_backward_hooks, non_full_backward_hooks = [], []

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\speechbrain\lobes\models\transformer\TransformerASR.py:348](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/speechbrain/lobes/models/transformer/TransformerASR.py:348), in EncoderWrapper.forward(self, x, wav_lens)
    346 def forward(self, x, wav_lens=None):
    347     """ Processes the input tensor x and returns an output tensor."""
--> 348     x = self.transformer.encode(x, wav_lens)
    349     return x

File [c:\Users\francisw\Documents\repos\speech-to-text\.venv\Lib\site-packages\speechbrain\lobes\models\transformer\TransformerASR.py:301](file:///C:/Users/francisw/Documents/repos/speech-to-text/.venv/Lib/site-packages/speechbrain/lobes/models/transformer/TransformerASR.py:301), in TransformerASR.encode(self, src, wav_len)
    298     pos_embs_source = self.positional_encoding(src)
    300 elif self.positional_encoding_type == "fixed_abs_sine":
--> 301     src = src + self.positional_encoding(src)
    302     pos_embs_source = None
    304 encoder_out, _ = self.encoder(
    305     src=src,
    306     src_key_padding_mask=src_key_padding_mask,
    307     pos_embs=pos_embs_source,
    308 )

RuntimeError: The size of tensor a (140590) must match the size of tensor b (2500) at non-singleton dimension 1

waveform = asr_model.load_audio("my/local/path/file.wav")
batch = waveform.unsqueeze(0)
rel_length = torch.tensor([1.0])
print(batch.shape)

torch.Size([1, 89977000])