[Bug] 'GPT2InferenceModel' object has no attribute 'generate'

[ErimatOesteRP]

Describe the bug

uv run .\teste1.py
Arquivo WAV: Sample rate=24000, Channels=1

tts_models/multilingual/multi-dataset/xtts_v2 is already downloaded.
Using model: xtts
GPT2InferenceModel has generative capabilities, as prepare_inputs_for_generation is explicitly defined. However, it doesn't directly inherit from GenerationMixin. From 👉v4.50👈 onwards, PreTrainedModel will NOT inherit from GenerationMixin, and this model will lose the ability to call generate and other related functions.

• If you're using trust_remote_code=True, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes
• If you are the owner of the model architecture code, please modify your model class such that it inherits from GenerationMixin (after PreTrainedModel, otherwise you'll get an exception).
• If you are not the owner of the model architecture class, please contact the model code owner to update it.

Text splitted to sentences.
['esse é um teste de clonagem de voz']
Traceback (most recent call last):
File "D:\bkp_hd\Projetos\python\TTS_02\teste1.py", line 22, in
tts.tts_to_file(text=text, speaker_wav=wav_path, language="pt", file_path=output_path)
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\api.py", line 334, in tts_to_file
wav = self.tts(
^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\api.py", line 276, in tts
wav = self.synthesizer.tts(
^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\utils\synthesizer.py", line 386, in tts
outputs = self.tts_model.synthesize(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\tts\models\xtts.py", line 419, in synthesize
return self.full_inference(text, speaker_wav, language, **settings)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\torch\utils_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\tts\models\xtts.py", line 488, in full_inference
return self.inference(
^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\torch\utils_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\tts\models\xtts.py", line 541, in inference
gpt_codes = self.gpt.generate(
^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\TTS\tts\layers\xtts\gpt.py", line 590, in generate
gen = self.gpt_inference.generate(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\bkp_hd\Projetos\python\TTS_02.venv\Lib\site-packages\torch\nn\modules\module.py", line 1940, in getattr
raise AttributeError(
AttributeError: 'GPT2InferenceModel' object has no attribute 'generate'

Atualização do modelo para suportar transformers 4.50+

To Reproduce

rom TTS.api import TTS
from TTS.tts.configs.xtts_config import XttsConfig
from TTS.tts.models.xtts import XttsAudioConfig, XttsArgs
from TTS.config.shared_configs import BaseDatasetConfig
import torch
import soundfile as sf

Adiciona os globals à lista de permitidos

torch.serialization.add_safe_globals([XttsConfig, XttsAudioConfig, BaseDatasetConfig, XttsArgs])

Verificar arquivo WAV de referência

wav_path = "D:/bkp_hd/Projetos/python/TTS_02/temp/referencia.wav"
data, sample_rate = sf.read(wav_path)
print(f"Arquivo WAV: Sample rate={sample_rate}, Channels={data.shape[1] if data.ndim > 1 else 1}")

Inicializar modelo TTS

tts = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", progress_bar=True, gpu=False)

Gerar áudio

text = "esse é um teste de clonagem de voz"
output_path = "D:/bkp_hd/Projetos/python/TTS_02/temp/output.wav"
tts.tts_to_file(text=text, speaker_wav=wav_path, language="pt", file_path=output_path)

print(f"Áudio gerado com sucesso: {output_path}")

Expected behavior

No response

Logs

Environment

uv pip show TTS transformers soundfile
Name: soundfile
Version: 0.13.1
Location: D:\bkp_hd\Projetos\python\TTS_02\.venv\Lib\site-packages
Requires: cffi, numpy
Required-by: librosa, trainer, tts
---
Name: transformers
Version: 4.52.1
Location: D:\bkp_hd\Projetos\python\TTS_02\.venv\Lib\site-packages
Requires: filelock, huggingface-hub, numpy, packaging, pyyaml, regex, requests, safetensors, tokenizers, tqdm
Required-by: tts
---
Required-by: tts
---
Name: tts
Version: 0.22.0

Additional context

No response

[eginhard]

This is fixed in our fork (available via pip install coqui-tts). This repo is not maintained anymore.

[ErimatOesteRP]

Thank you very much

[wapping]

Downgrade the transformers. It works for me.
pip install transformers==4.33.0

[ErimatOesteRP]

I will downgrade, tank you very much, it worked now...

[eginhard]

I'd recommend to use the fork (pip install coqui-tts) instead. It supports newer transformer versions and has many other bugs fixed.

[ibaoger]

看了项目的 requirements.txt 内容，这个问题出现的根本原因是 TTS 没有限制依赖项的版本，导致安装了最新版本的依赖，而TTS本身又不兼容新版本，最好的办法是限制每一个依赖项的版本，一劳永逸的解决问题。

[aspirant2018]

看了项目的 requirements.txt 内容，这个问题出现的根本原因是 TTS 没有限制依赖项的版本，导致安装了最新版本的依赖，而TTS本身又不兼容新版本，最好的办法是限制每一个依赖项的版本，一劳永逸的解决问题。

I translated this to English because it helped me alot to understand the issue - i used Chatgpt

Translation to English:
'After reviewing the project's requirements.txt, the root cause of the issue is that TTS does not specify version constraints for its dependencies. As a result, the latest versions of the dependencies are installed, which are not compatible with TTS itself. The best solution is to explicitly specify the version of each dependency to resolve the problem once and for all.'

[aspirant2018]

This list of packages with the correpondant versions that helped me to run the inference of TTS correctly:

absl-py==2.3.0
aiohappyeyeballs==2.6.1
aiohttp==3.12.13
aiosignal==1.3.2
annotated-types==0.7.0
anyascii==0.3.2
attrs==25.3.0
audioread==3.0.1
babel==2.17.0
bangla==0.0.5
blinker==1.9.0
blis==1.2.1
bnnumerizer==0.0.2
bnunicodenormalizer==0.1.7
catalogue==2.0.10
certifi==2025.6.15
cffi==1.17.1
charset-normalizer==3.4.2
click==8.2.1
cloudpathlib==0.21.1
colorama==0.4.6
confection==0.1.5
contourpy==1.3.2
coqpit==0.0.17
cycler==0.12.1
cymem==2.0.11
Cython==3.1.2
dateparser==1.1.8
decorator==5.2.1
docopt==0.6.2
einops==0.8.1
encodec==0.1.1
filelock==3.18.0
Flask==3.1.1
fonttools==4.58.4
frozenlist==1.7.0
fsspec==2025.5.1
g2pkk==0.1.2
grpcio==1.73.0
gruut==2.2.3
gruut-ipa==0.13.0
gruut_lang_de==2.0.1
gruut_lang_en==2.0.1
gruut_lang_es==2.0.1
gruut_lang_fr==2.0.2
hangul-romanize==0.1.0
huggingface-hub==0.33.0
idna==3.10
inflect==7.5.0
itsdangerous==2.2.0
jamo==0.4.1
jieba==0.42.1
Jinja2==3.1.6
joblib==1.5.1
jsonlines==1.2.0
kiwisolver==1.4.8
langcodes==3.5.0
language_data==1.3.0
lazy_loader==0.4
librosa==0.11.0
llvmlite==0.44.0
marisa-trie==1.2.1
Markdown==3.8.2
markdown-it-py==3.0.0
MarkupSafe==3.0.2
matplotlib==3.10.3
mdurl==0.1.2
more-itertools==10.7.0
mpmath==1.3.0
msgpack==1.1.1
multidict==6.5.0
murmurhash==1.0.13
networkx==2.8.8
nltk==3.9.1
num2words==0.5.14
numba==0.61.2
numpy==1.26.4
packaging==25.0
pandas==1.5.3
pillow==11.0.0
platformdirs==4.3.8
pooch==1.8.2
preshed==3.0.10
propcache==0.3.2
protobuf==6.31.1
psutil==7.0.0
pycparser==2.22
pydantic==2.11.7
pydantic_core==2.33.2
Pygments==2.19.1
pynndescent==0.5.13
pyparsing==3.2.3
pypinyin==0.54.0
pysbd==0.3.4
python-crfsuite==0.9.11
python-dateutil==2.9.0.post0
pytz==2025.2
PyYAML==6.0.2
regex==2024.11.6
requests==2.32.4
rich==14.0.0
safetensors==0.5.3
scikit-learn==1.7.0
scipy==1.15.3
shellingham==1.5.4
six==1.17.0
smart-open==7.1.0
soundfile==0.13.1
soxr==0.5.0.post1
spacy==3.8.7
spacy-legacy==3.0.12
spacy-loggers==1.0.5
srsly==2.5.1
SudachiDict-core==20250515
SudachiPy==0.6.10
sympy==1.13.1
tensorboard==2.19.0
tensorboard-data-server==0.7.2
thinc==8.3.4
threadpoolctl==3.6.0
tokenizers==0.13.3
torch==2.5.1+cu121
torchaudio==2.5.1+cu121
torchvision==0.20.1+cu121
tqdm==4.67.1
trainer==0.0.36
transformers==4.33.0
TTS==0.22.0
typeguard==4.4.4
typer==0.16.0
typing-inspection==0.4.1
typing_extensions==4.14.0
tzdata==2025.2
tzlocal==5.3.1
umap-learn==0.5.7
Unidecode==1.4.0
urllib3==2.5.0
wasabi==1.1.3
weasel==0.4.1
Werkzeug==3.1.3
wrapt==1.17.2
yarl==1.20.1

[eginhard]

@ibaoger @aspirant2018 As mentioned above, I'd recommend to just install the fork instead with pip install coqui-tts. It has been updated to work with recent versions of packages out of the box and includes many other bug fixes.

[aspirant2018]

@eginhard yes, i have seen your comment and your recommendation. I will use the forked package sooner or later. Thank you again