windows系统部署问题

[Jandown]

使用gradio生成，卡在Loading HiFT

[load_quantize_encoder] start. model_path='ckpt/speech_tokenizer'
Configured for 24kHz frontend.
2025-12-12 09:52:10.6152300 [W:onnxruntime:, transformer_memcpy.cc:74 onnxruntime::MemcpyTransformer::ApplyImpl] 2 Memcpy nodes are added to the graph torch_jit for CUDAExecutionProvider. It might have negative impact on performance (including unable to run CUDA graph). Set session_options.log_severity_level=1 to see the detail logs before this message.
Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 5.51s/it]
Loading HiFT model from ckpt/hift/hift.pt on cuda...

还有，希望能优化下对win的支持，否则坑太多

[Cocii]

Thank you for your feedback!
We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

[Jandown]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

暂时没有其他的，还没跑起来，所以只有这些，感谢回复

[tool2d]

需要9G显存, 目前至少需要12G显存显卡, 才能跑起来.

[Jandown]

需要9G显存, 目前至少需要12G显存显卡, 才能跑起来.

跟显存没关系，我是RTX5090

[PierrunoYT]

For me it doesnt work because of the Pyini and WeTextProcessing

[Jandown]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

是不是gradio的问题， 我直接运行推理 glmtts_inference.py 是正常的，使用gradio就卡在Loading HiFT model from ckpt/hift/hift.pt on cuda...

[Jandown]

For me it doesnt work because of the Pyini and WeTextProcessing

conda install -c conda-forge pynini=2.1.5
pip install WeTextProcessing==1.0.3 --no-deps

[gluttony-10]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

是不是gradio的问题， 我直接运行推理 glmtts_inference.py 是正常的，使用gradio就卡在Loading HiFT model from ckpt/hift/hift.pt on cuda...

#24

[Jandown]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

是不是gradio的问题， 我直接运行推理 glmtts_inference.py 是正常的，使用gradio就卡在Loading HiFT model from ckpt/hift/hift.pt on cuda...

#24

感谢十字鱼大佬的pr，我下午也发现了，是gradio版本的问题，已经解决了。

[gluttony-10]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

是不是gradio的问题， 我直接运行推理 glmtts_inference.py 是正常的，使用gradio就卡在Loading HiFT model from ckpt/hift/hift.pt on cuda...

#24

感谢十字鱼大佬的pr，我下午也发现了，是gradio版本的问题，已经解决了。

这个确实不好发现，没有报错的。导致前几天我跟onnxruntime搏斗了好久。现在该去修GLM-ASR了

[Jandown]

Thank you for your feedback! We will resolve it as soon as possible. Are there any other issues you are experiencing with the Windows client?

是不是gradio的问题， 我直接运行推理 glmtts_inference.py 是正常的，使用gradio就卡在Loading HiFT model from ckpt/hift/hift.pt on cuda...

#24

感谢十字鱼大佬的pr，我下午也发现了，是gradio版本的问题，已经解决了。

这个确实不好发现，没有报错的。导致前几天我跟onnxruntime搏斗了好久。现在该去修GLM-ASR了

因为我下午测试GLM-ASR，做了一个简单的webui，用gradio6也是卡住，直接推理就正常，当时想到是不是gradio版本的问题，结果换了5就正常了，然后试了试glm-tts也降级到5，结果就跑起来了。。。