I have been using the A100 to train the ISNet model. However, I encountered a problem with CUDA incompatibility. To resolve this issue, I spent some time upgrading the CUDA-related packages and conducting compatibility checks. Now, ISNet can run on newer architecture GPUs like the A100/4090. I have exported the upgraded environment configuration to share with you, hoping to help more people avoid the pain of environment upgrading.
The environment configuration files I am providing are compatible with CUDA 11.8. Using this environment, ISNet can run on GPUs with Ampere architecture and earlier, such as the 30 series cards, 40 series cards, A100, A10, etc. Except for the H100, which requires a CUDA 12+ environment, CUDA 11.8 currently supports the vast majority of Nvidia graphics cards.
In addition, I have provided three configuration methods for the CUDA 11.8 environment in my GitHub. If you are unable to configure the environment through yaml and txt, please go to my github and download the Conda environment compression package through onedrive, and directly extract it for use.
requirements_cu118.txt
I have been using the A100 to train the ISNet model. However, I encountered a problem with CUDA incompatibility. To resolve this issue, I spent some time upgrading the CUDA-related packages and conducting compatibility checks. Now, ISNet can run on newer architecture GPUs like the A100/4090. I have exported the upgraded environment configuration to share with you, hoping to help more people avoid the pain of environment upgrading.
The environment configuration files I am providing are compatible with CUDA 11.8. Using this environment, ISNet can run on GPUs with Ampere architecture and earlier, such as the 30 series cards, 40 series cards, A100, A10, etc. Except for the H100, which requires a CUDA 12+ environment, CUDA 11.8 currently supports the vast majority of Nvidia graphics cards.
In addition, I have provided three configuration methods for the CUDA 11.8 environment in my GitHub. If you are unable to configure the environment through yaml and txt, please go to my github and download the Conda environment compression package through onedrive, and directly extract it for use. requirements_cu118.txt