made an account just to chime in positively about the excellent support
can confirm was getting the same error on a 1080Ti and recompiling xformers from source fixed it as detailed in the reddit post linked by @GeneralAwareness
As of my last knowledge update in September 2021, Torch is a scientific computing framework that primarily focuses on machine learning tasks, including deep learning. Torch doesn’t have strict GPU support restrictions itself; rather, its compatibility with specific GPUs largely depends on the underlying libraries it uses, such as CUDA and cuDNN.
At that time 3d scan, Torch was commonly used with NVIDIA GPUs due to its integration with CUDA, NVIDIA’s parallel computing platform. This allowed deep learning operations to be accelerated on compatible NVIDIA GPUs. The specific GPUs supported by Torch are generally determined by the version of CUDA and cuDNN that Torch is compiled against.
For the most accurate and up-to-date information regarding Torch’s GPU compatibility, I recommend checking the official Torch repository or documentation. Additionally, consider checking if there have been any updates or changes to Torch’s GPU support beyond September 2021, as my knowledge is based on information available up to that point.