最后一步python gradio_app.py报错了

[lgkt]

(omost) PS D:\ai\Omost> python gradio_app.py D:\ai\Omost\lib_omost\pipeline.py:64: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requiresgrad(True), rather than torch.tensor(sourceTensor). alphas_cumprod = torch.tensor(np.cumprod(alphas, axis=0), dtype=torch.float32) Unload to CPU: AutoencoderKL Unload to CPU: CLIPTextModel Unload to CPU: CLIPTextModel Unload to CPU: UNet2DConditionModel Unused kwargs: ['_load_in_4bit', '_load_in_8bit', 'quant_method']. These kwargs are not used in <class 'transformers.utils.quantization_config.BitsAndBytesConfig'>. Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:03<00:00, 1.55s/it] WARNING:root:Some parameters are on the meta device device because they were offloaded to the cpu. Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. WARNING:accelerate.big_modeling:You shouldn't move a model that is dispatched using accelerate hooks. Traceback (most recent call last): File "D:\ai\Omost\gradio_app.py", line 87, in memory_management.unload_all_models(llm_model) File "D:\ai\Omost\lib_omost\memory_management.py", line 67, in unload_all_models return load_models_to_gpu([]) File "D:\ai\Omost\lib_omost\memory_management.py", line 42, in load_models_to_gpu m.to(cpu) File "C:\Users\lgkt\AppData\Roaming\Python\Python310\site-packages\accelerate\big_modeling.py", line 455, in wrapper raise RuntimeError("You can't move a model that has some modules offloaded to cpu or disk.") RuntimeError: You can't move a model that has some modules offloaded to cpu or disk.

[xhoxye]

如果你使用的是原版代码，可以尝试其他pr，或者自己修改 If you're using the original code, you can try other PRs, or modify it yourself

https://github.com/xhoxye/Omost_CN

[lgkt]

是原版，不太会coding，不知道咋改……我只是想用这个omost，hf上的space时间太有限了。如果是你的pr，估计也会有这个问题吧 @xhoxye

[lgkt]

"D:\ai\Omost\lib_omost\memory_management.py"这个文件我倒是找到了，但是咋改呢 @lllyasviel 敏神

[xhoxye]

@lgkt 我自己使用是没有这个问题，不知道是不是原版的问题，因为它只是一个演示，你的电脑配置是什么？

https://github.com/lllyasviel/Omost/pull/81

这个是我的pr

[lgkt]

我3070 8G VRAM。我看你的pr，那个内存管理文件和原版是一样的 @xhoxye

[takishun0326]

I resolved the issue by specifying a particular GPU in the code.

There were others experiencing a similar problem, but I believe the cause was that their GPU did not meet the 8GB memory requirement, resulting in that GPU being selected. In such cases, you can specify a GPU with over 8GB of VRAM in gradio_app.py using os.environ["CUDA_VISIBLE_DEVICES"] = "0" to select the 0th GPU or by using some other method to resolve the issue.