Awesome. I'm installing on Ubuntu 22.04 right now.
Ran into a few errors with the default instructions related to CUDA version mismatches with my nvidia driver. Now I'm trying without conda at all. Made a venv. I upgraded to the latest that Ubuntu provides and then downloaded and installed the appropriate CUDA from [1].
That got me farther. Then ran into the fact that the xformers binaries I had in my earlier attempts is now incompatible with my current drivers and CUDA, so rebuiding that one. I'm in the 30-minute compile, but did the `pip install ninja` as recommended by [2] and it's running on a few of my 32 threads now. Ope! Done in 5 mins. Test info from `python -m xformers.info` looks good.
Damn still hitting CUDA out of memory issues. I knew I should have bought a bigger GPU back in 2017. Everyone says I have to downgrade pytorch to 1.12.1 for this to not happen. But oh dang that was compiled with a different cuda, oh groan. Maybe I should get conda to work afterall.
`torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 30.00 MiB (GPU 0; 5.93 GiB total capacity; 5.62 GiB already allocated; 15.44 MiB free; 5.67 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF`
Guess I better go read those docs... to be continued.
GeForce GTX 1060 6GB, purchased literally 5 years ago. It worked with an optimized stable diffusion 1.0 so I was hopeful here. If I want to run these models going forward I guess I need something slightly more serious, eh?
Ran into a few errors with the default instructions related to CUDA version mismatches with my nvidia driver. Now I'm trying without conda at all. Made a venv. I upgraded to the latest that Ubuntu provides and then downloaded and installed the appropriate CUDA from [1].
That got me farther. Then ran into the fact that the xformers binaries I had in my earlier attempts is now incompatible with my current drivers and CUDA, so rebuiding that one. I'm in the 30-minute compile, but did the `pip install ninja` as recommended by [2] and it's running on a few of my 32 threads now. Ope! Done in 5 mins. Test info from `python -m xformers.info` looks good.
Damn still hitting CUDA out of memory issues. I knew I should have bought a bigger GPU back in 2017. Everyone says I have to downgrade pytorch to 1.12.1 for this to not happen. But oh dang that was compiled with a different cuda, oh groan. Maybe I should get conda to work afterall.
`torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 30.00 MiB (GPU 0; 5.93 GiB total capacity; 5.62 GiB already allocated; 15.44 MiB free; 5.67 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF`
Guess I better go read those docs... to be continued.
[1] https://developer.nvidia.com/cuda-downloads?target_os=Linux&...
[2] https://github.com/facebookresearch/xformers