AssertionError: Torch not compiled with CUDA enabled

[TheCompAce]

Well I got passed the Type Error, and ran into this, I looked it up on google, and the only solutions I saw was to change the model to not include it. It seems to be erroring out on line "background_mapping," for "video_dataset.py"

Traceback (most recent call last): File "T:\Tensorflow\Programs\depplearn\tools\Text2LIVE\train_video.py", line 102, in train_model(config) File "T:\Tensorflow\Programs\depplearn\tools\Text2LIVE\train_video.py", line 28, in train_model dataset = AtlasDataset(config) File "T:\Tensorflow\Programs\depplearn\tools\Text2LIVE\datasets\video_dataset.py", line 32, in init self.original_video = load_video( File "C:\Users\Tha Killa.conda\envs\text2live\lib\site-packages\torch\cuda_init_.py", line 208, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

[Meatfucker]

This is an issue with their package set. remove your existing conda enviroment and create a new one with this. conda create --name text2live python=3.9 pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge

Also, I OOM'd about 75% of the way through a training session with a 24gb 3090, so the requirement of 32gb seems to be accurate.

[TheCompAce]

Well I tried "conda create --name text2live python=3.9 pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge" (I also removed torch from the requirements.txt file) But still the same.

[Meatfucker]

You may need to adjust the Cuda version depending on your card and its cuda capabilities. The pytorch website has an archive of old commands to pull in older versions. Find out what cuda version your card supports and go from there. Also be sure to completely remove your previous conda enviroment as existing packages will gunk up solving the dependency chain.