weights_only problem when using XTTS

[Nikanoru]

Sorry it is me again! Running with XTTS always gives me the following error.

epub2tts VITAEPATRUMBOOK3.txt --engine xtts --speaker "Damien Black" --end 2

Loading model: /home/david/.local/share/tts/tts_models--multilingual--multi-dataset--xtts_v2 Traceback (most recent call last): File "/home/david/.local/bin/epub2tts", line 8, in sys.exit(main()) File "/home/david/.local/lib/python3.10/site-packages/epub2tts.py", line 1239, in main mybook.read_book( File "/home/david/.local/lib/python3.10/site-packages/epub2tts.py", line 941, in read_book map_result = list(map(process_book_chapter, chapter_job_que)) File "/home/david/.local/lib/python3.10/site-packages/epub2tts.py", line 353, in process_book_chapter tts_engine = dat['config']'engine_cl' File "/home/david/.local/lib/python3.10/site-packages/epub2tts.py", line 178, in init tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(self.device) File "/home/david/.local/lib/python3.10/site-packages/TTS/api.py", line 75, in init self.load_tts_model_by_name(model_name, gpu) File "/home/david/.local/lib/python3.10/site-packages/TTS/api.py", line 177, in load_tts_model_by_name self.synthesizer = Synthesizer( File "/home/david/.local/lib/python3.10/site-packages/TTS/utils/synthesizer.py", line 112, in init self._load_tts_from_dir(model_dir, use_cuda) File "/home/david/.local/lib/python3.10/site-packages/TTS/utils/synthesizer.py", line 167, in _load_tts_from_dir self.tts_model.load_checkpoint(config, checkpoint_dir=model_dir, eval=True) File "/home/david/.local/lib/python3.10/site-packages/TTS/tts/models/xtts.py", line 778, in load_checkpoint checkpoint = self.get_compatible_checkpoint_state_dict(model_path) File "/home/david/.local/lib/python3.10/site-packages/TTS/tts/models/xtts.py", line 719, in get_compatible_checkpoint_state_dict checkpoint = load_fsspec(model_path, map_location=torch.device("cpu"))["model"] File "/home/david/.local/lib/python3.10/site-packages/TTS/utils/io.py", line 54, in load_fsspec return torch.load(f, map_location=map_location, **kwargs) File "/home/david/.local/lib/python3.10/site-packages/torch/serialization.py", line 1470, in load raise pickle.UnpicklingError(_get_wo_message(str(e))) from None _pickle.UnpicklingError: Weights only load failed. This file can still be loaded, to do so you have two options, do those steps only if you trust the source of the checkpoint. (1) In PyTorch 2.6, we changed the default value of the weights_only argument in torch.load from False to True. Re-running torch.load with weights_only set to False will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with weights_only=True please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL TTS.tts.configs.xtts_config.XttsConfig was not an allowed global by default. Please use torch.serialization.add_safe_globals([XttsConfig]) or the torch.serialization.safe_globals([XttsConfig]) context manager to allowlist this global if you trust this class/function. Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.

[jakowisp]

This is caused by a security update in how pytorch works. I resolved this by the following method.

1. use git to pull https://github.com/coqui-ai/TTS.git 2)modifly TTS/utils/io.py
   • modify the 2 torch.load lines. as shown below. This will enable the previous versions of coqui functionality, but it does open a security hole. return torch.load(f, ,weights_only=False, map_location=map_location, **kwargs) return torch.load(f, weights_only=False, map_location=map_location, **kwargs)
2. pip install .[all] This installs the TTS library that we have modified for use by epub2tts.