#7295 Tensor.item() cannot be called on meta tensors,Traceback (most recent call last): File

85.99**2 Posted at: 1 hour ago

Tensor.item() cannot be called on meta tensors,Traceback (most recent call last):
File "videotrans\process\qwen_tts.py", line 125, in qwen3tts_fun
File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\utils\_contextlib.py", line 116, in decorate_context

return func(*args, **kwargs)

File "qwen_tts\inference\qwen3_tts_model.py", line 603, in generate_voice_clone

talker_codes_list, _ = self.model.generate(

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\utils\_contextlib.py", line 116, in decorate_context

return func(*args, **kwargs)

File "qwen_tts\core\models\modeling_qwen3_tts.py", line 2272, in generate

talker_result = self.talker.generate(

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\utils\_contextlib.py", line 116, in decorate_context

return func(*args, **kwargs)

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\generation\uti
......
nesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\generation\utils.py", line 2789, in _sample

outputs = model_forward(**model_inputs, return_dict=True)

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl

return self._call_impl(*args, **kwargs)

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\nn\modules\module.py", line 1762, in _call_impl

return forward_call(*args, **kwargs)

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\utils\generic.py", line 918, in wrapper

output = func(self, *args, **kwargs)

File "qwen_tts\core\models\modeling_qwen3_tts.py", line 1671, in forward

predictor_result = self.code_predictor.generate(

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\utils\_contextlib.py", line 116, in decorate_context

return func(*args, **kwargs)

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\generation\utils.py", line 2566, in generate

result = decoding_method(

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\generation\utils.py", line 2781, in _sample

while self._has_unfinished_sequences(this_peer_finished, synced_gpus, device=input_ids.device):

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\transformers4576\generation\utils.py", line 2599, in _has_unfinished_sequences

elif this_peer_finished:

File "C:\Users\enesg\Downloads\win-pyvideotrans-v4.12-0912\_internal\torch\_meta_registrations.py", line 7088, in meta_local_scalar_dense

raise RuntimeError("Tensor.item() cannot be called on meta tensors")

RuntimeError: Tensor.item() cannot be called on meta tensors

=
system:Windows-10-10.0.26200-SP0
version:v4.12
frozen:True
language:en_US
root_dir:C:/Users/enesg/Downloads/win-pyvideotrans-v4.12-0912

#1 DeepSeek reply at 1 hour ago

Core cause: The Qwen3-TTS (built-in) channel failed during voice cloning. The error Tensor.item() cannot be called on meta tensors typically indicates the model was loaded incompletely/corrupted (a meta tensor is a placeholder, not real weights) — most commonly due to a failed or partial model download from HuggingFace, or insufficient VRAM during load.

Try the following in order:

Delete and re-download the Qwen3-TTS model (most likely fix). The model files are likely incomplete/corrupted.
- Delete the folder: 软件目录/models/models--Qwen--Qwen3-TTS-12Hz-0.6B-Base
- Delete the folder: 软件目录/models/models--Qwen--Qwen3-TTS-12Hz-0.6B-CustomVoice
- Restart the software so it re-downloads the models (use a stable VPN/proxy if downloads fail; the built-in mirror hf-mirror.com will be used automatically without a proxy).
- Manual download addresses:
  - https://huggingface.co/Qwen/Qwen3-TTS-12Hz-0.6B-Base/tree/main
  - https://huggingface.co/Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice/tree/main
Verify reference audio length. Clone reference audio must be 3–10s. In 菜单 -> 工具 -> 高级选项 -> 语音识别参数, set:
- 最短语音持续(毫秒) ≥ 3000
- 最长语音持续(秒) between 6–10
Check VRAM. Qwen3-TTS uses GPU first. If you have a small NVIDIA GPU or AMD/Intel GPU, try a different TTS channel (e.g. Edge-TTS or F5-TTS), or run with less demanding models.
Re-download the complete package if patched. If you applied a 260MB patch over an old install, replace with the full 2.7GB package to ensure no library/model files are missing.
If the error persists after steps 1–4, temporarily switch the dubbing channel to Edge-TTS to confirm the rest of the pipeline works, then re-test Qwen3-TTS.

#7295 Tensor.item() cannot be called on meta tensors,Traceback (most recent call last): File

1 Replies

请查阅相关文档：

Post Your Reply

Related resource