Torch not compiled with CUDA enabled

[TanvirHafiz]

it says "Torch not compiled with CUDA enabled" how do i enable CUDA and compile torch??? or am I doing something wrong!

I get this

C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torchvision\io\image.py:13: UserWarning: Failed to load image Python extension: Could not find module 'C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\Lib\site-packages\torchvision\image.pyd' (or one of its dependencies). Try using the full path with constructor syntax. warn(f"Failed to load image Python extension: {e}") No CUDA runtime is found, using CUDA_HOME='C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1' C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\mmcv__init.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details. warnings.warn( Initializing BaseSegmenter to cuda:3 Traceback (most recent call last): File "G:\Track-Anything\app.py", line 385, in model = TrackingAnything(SAM_checkpoint, xmem_checkpoint, e2fgvi_checkpoint,args) File "G:\Track-Anything\track_anything.py", line 18, in init self.samcontroler = SamControler(self.sam_checkpoint, args.sam_model_type, args.device) File "G:\Track-Anything\tools\interact_tools.py", line 37, in init self.sam_controler = BaseSegmenter(SAM_checkpoint, model_type, device) File "G:\Track-Anything\tools\base_segmenter.py", line 26, in init__ self.model.to(device=self.device) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 907, in to return self._apply(convert) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 578, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 578, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 578, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 601, in _apply param_applied = fn(param) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 905, in convert return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\cuda__init__.py", line 210, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

but I have cuda 12.1 and 11.7 installed

[TanvirHafiz]

reinstalled torch frest now digging this issue

C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\mmcv__init.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details. warnings.warn( download checkpoints ...... download successfully! Initializing BaseSegmenter to cuda:3 Traceback (most recent call last): File "G:\Track-Anything\app.py", line 385, in model = TrackingAnything(SAM_checkpoint, xmem_checkpoint, e2fgvi_checkpoint,args) File "G:\Track-Anything\track_anything.py", line 18, in init self.samcontroler = SamControler(self.sam_checkpoint, args.sam_model_type, args.device) File "G:\Track-Anything\tools\interact_tools.py", line 37, in init self.sam_controler = BaseSegmenter(SAM_checkpoint, model_type, device) File "G:\Track-Anything\tools\base_segmenter.py", line 26, in init__ self.model.to(device=self.device) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1149, in to return self._apply(convert) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 801, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 801, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 801, in _apply module._apply(fn) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 824, in _apply param_applied = fn(param) File "C:\Users\Tanvir\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1147, in convert return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking) RuntimeError: CUDA error: invalid device ordinal CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

[Monkiesh]

Same issue here.

[mfkeles]

same here any chance this was resolved?

[jin-eld]

The error message is misleading, I think the issue is that app.py hardcodes args.device = "cuda:3" which will fail if you have only one card. Try removing that line, in this case the --device command line parameter will actually work again and will also use cuda:0 as default, which should be fine for most users.

[aboharbf]

Same issue, even when commenting out the cuda:3 line

Torch not compiled with CUDA enabled
  File "C:\Track-Anything\tools\base_segmenter.py", line 26, in __init__
    self.model.to(device=self.device)
  File "C:\Track-Anything\tools\interact_tools.py", line 37, in __init__
    self.sam_controler = BaseSegmenter(SAM_checkpoint, model_type, device)
  File "C:\Track-Anything\track_anything.py", line 18, in __init__
    self.samcontroler = SamControler(self.sam_checkpoint, args.sam_model_type, args.device)
  File "C:\Track-Anything\app.py", line 385, in <module>
    model = TrackingAnything(SAM_checkpoint, xmem_checkpoint, e2fgvi_checkpoint,args)
AssertionError: Torch not compiled with CUDA enabled

[mfkeles]

it’s not the same actually you have to install torch that matches your cuda version.

[aboharbf]

I was looking at other posts about this problem, and after putting

        import torch

        print("Torch version:",torch.__version__)

        print("Is CUDA enabled?",torch.cuda.is_available())

I find

Torch version: 2.0.1+cpu
Is CUDA enabled? False

I use a 3080, so not sure why the CPU version is being used. I was under the impression the pip requirements would install CUDA but now I'm getting the package straight from Nvidia via .exe

[mfkeles]

you have to conda install torch with the cuda version that it is compatible. you got torch 2.0.1 installed i think that requires cuda 12 at least

[aboharbf]

Seems like I still have the same problem. Even deleted the env and clean installed

I'll have to dig into why CUDA may not be turning on despite the env seemingly 'seeing' it.

[aboharbf]

Update: I got it working by following the procedure on this page: https://www.educative.io/answers/how-to-resolve-torch-not-compiled-with-cuda-enabled. Had to uninstall Torch and reinstall using instructions here: https://pytorch.org/get-started/locally/

[samadbarrikhojasteh]

Please change "cuda:3" to "cuda:0" in app.py file (line 380) then install gradio pip install gradio==3.39.0.

[TanvirHafiz]

(venv) F:\Track-Anything>python app.py --device cuda:0 F:\Track-Anything\venv\lib\site-packages\gradio_client\documentation.py:103: UserWarning: Could not get documentation group for <class 'gradio.mix.Parallel'>: No known documentation group for module 'gradio.mix' warnings.warn(f"Could not get documentation group for {cls}: {exc}") F:\Track-Anything\venv\lib\site-packages\gradio_client\documentation.py:103: UserWarning: Could not get documentation group for <class 'gradio.mix.Series'>: No known documentation group for module 'gradio.mix' warnings.warn(f"Could not get documentation group for {cls}: {exc}") Initializing BaseSegmenter to cuda:0 Traceback (most recent call last): File "F:\Track-Anything\app.py", line 385, in model = TrackingAnything(SAM_checkpoint, xmem_checkpoint, e2fgvi_checkpoint,args) File "F:\Track-Anything\track_anything.py", line 18, in init self.samcontroler = SamControler(self.sam_checkpoint, args.sam_model_type, args.device) File "F:\Track-Anything\tools\interact_tools.py", line 37, in init self.sam_controler = BaseSegmenter(SAM_checkpoint, model_type, device) File "F:\Track-Anything\tools\base_segmenter.py", line 26, in init self.model.to(device=self.device) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 1152, in to return self._apply(convert) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 802, in _apply module._apply(fn) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 802, in _apply module._apply(fn) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 802, in _apply module._apply(fn) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 825, in _apply param_applied = fn(param) File "F:\Track-Anything\venv\lib\site-packages\torch\nn\modules\module.py", line 1150, in convert return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking) File "F:\Track-Anything\venv\lib\site-packages\torch\cuda__init__.py", line 293, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

changed GPU to 3090. all the other programs that uses CUDA works. except this one. removed the line with cuda:3, installed gradio 3.39, sti;; same error