[FastPitch/Pytorch] size mismatch for embedding.weight

[Truong-Thanh-Quang]

I use my own pretrained Tacotron2 model FP16 to run extract mels. Error occured:

Traceback (most recent call last):
File "extract_mels.py", line 290, in
main()
File "extract_mels.py", line 199, in main
forward_is_infer=False, ema=False)
File "/workspace/fastpitch/inference.py", line 135, in load_and_setup_model
status += ' ' + str(model.load_state_dict(sd, strict=False))
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py", line 879, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for Tacotron2:
size mismatch for embedding.weight: copying a param with shape torch.Size([199, 512]) from checkpoint, the shape in current model is torch.Size([148, 512]).

Please help.

[kdorichev]

This error is due to the fact that by default the length of internal dicttionary used which is 148, while your Tacotron2 model is pretrained with 199 symbols in the dictionary.

Try to pass the argument --n-symbols.

[Truong-Thanh-Quang]

Thank @kdorichev!