|
|
@@ -444,7 +444,7 @@ class LCBNet(nn.Module):
|
|
|
encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)
|
|
|
if isinstance(encoder_out, tuple):
|
|
|
encoder_out = encoder_out[0]
|
|
|
- pdb.set_trace()
|
|
|
+
|
|
|
ocr_list_new = [[x + 1 if x != 0 else x for x in sublist] for sublist in ocr_sample_list]
|
|
|
ocr = torch.tensor(ocr_list_new).to(device=kwargs["device"])
|
|
|
ocr_lengths = ocr.new_full([1], dtype=torch.long, fill_value=ocr.size(1)).to(device=kwargs["device"])
|