[Bug] Version 0.25.2 breaks the fairseq models

[DrewThomasson]

Describe the bug

Version 0.25.2 Is causing issues with fairseq
For Example when the eng.tar.gz is downloaded and extracted the G_100000.pth is being renamed to model.pth
This is causing issues as you might expect when the fairseq engine is looking for the G_100000.pth file and can't find it


To Reproduce

Try to use the "eng" fairseq model

This is what the zip should look like

wget https://dl.fbaipublicfiles.com/mms/tts/eng.tar.gz

Expected behavior

Expected output of that zip are the files

G_100000.pth config.json vocab.txt

But instead its getting

model.pth config.json vocab.txt

If I Rename the model.pth to G_100000.pth then the model loads fine, confirming my suspicion

They also have the same shard

Logs

Idk I forgot to get it but its pretty self-evident when you re-download the English fairseq model to use fully with coqui-tts

oh there was this

 bash

FileNotFoundError: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth'
Caught DependencyError: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth'
convert_ebook() Exception: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth' 

s]Traceback (most recent call last):
File "/Users/drew/ebook2audiobook/lib/functions.py", line 694, in convert_chapters_to_audio
params['tts'] = load_tts_api_cached(model_path)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/lib/functions.py", line 634, in load_tts_api_cached
tts = XTTS(model_path)
^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/api.py", line 96, in init
self.load_tts_model_by_name(model_name, vocoder_name, gpu=gpu)
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/api.py", line 226, in load_tts_model_by_name
self.synthesizer = Synthesizer(
^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/utils/synthesizer.py", line 109, in init
self._load_fairseq_from_dir(model_dir, use_cuda)
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/utils/synthesizer.py", line 157, in _load_fairseq_from_dir
self.tts_model.load_fairseq_checkpoint(self.tts_config, checkpoint_dir=model_dir, eval=True)
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/tts/models/vits.py", line 1605, in load_fairseq_checkpoint
new_chk = rehash_fairseq_vits_checkpoint(checkpoint_file)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/TTS/tts/utils/fairseq.py", line 7, in rehash_fairseq_vits_checkpoint
chk = torch.load(checkpoint_file, map_location=torch.device("cpu"), weights_only=is_pytorch_at_least_2_4())["model"]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/torch/serialization.py", line 1319, in load
with _open_file_like(f, "rb") as opened_file:
^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/torch/serialization.py", line 659, in _open_file_like
return _open_file(name_or_buffer, mode)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/drew/ebook2audiobook/python_env/lib/python3.12/site-packages/torch/serialization.py", line 640, in init
super().init(open(name, mode))
^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth'
Caught DependencyError: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth'
convert_ebook() Exception: [Errno 2] No such file or directory: '/Users/drew/ebook2audiobook/models/tts/tts_models--eng--fairseq--vits/G_100000.pth'
100%|█████████████████████████████████████████████████| 135M/135M [00:30<00:00, 8.23MiB/s]

### Environment

```shell
on m1 MAC pro apple silicon 16gb ram 

issues arrises when using 

`coqui-tts==0.25.2`

Additional context

If I revert back to coqui-tts==0.25.1 then the issue goes away so

yeah

[eginhard]

@ROBERT-MCDOWELL @DrewThomasson Thanks for spotting this issue quickly, will release a new version with a fix (#264) today. It will accept both G_100000.pth and model.pth so it should work for any already downloaded models. Also added a test for the fairseq models to avoid similar issues in the future.

[ROBERT-MCDOWELL]

Thanks @eginhard !!! True sometimes models have other names than model.pth....