зеркало из https://github.com/microsoft/DeepSpeed.git
[build] add compute_86 (#577)
RTX-30 series are compute_86 ``` python -c "import torch; print(torch.cuda.get_device_capability())" ``` This PR adds support for this compute capability. Reference: https://developer.nvidia.com/cuda-gpus Co-authored-by: Jeff Rasley <jerasley@microsoft.com>
This commit is contained in:
Родитель
ce363d0e06
Коммит
e8b126d986
|
@ -34,7 +34,7 @@ def installed_cuda_version():
|
|||
def get_default_compute_capatabilities():
|
||||
compute_caps = DEFAULT_COMPUTE_CAPABILITIES
|
||||
if installed_cuda_version()[0] >= 11:
|
||||
compute_caps += ";8.0"
|
||||
compute_caps += ";8.0;8.6"
|
||||
return compute_caps
|
||||
|
||||
|
||||
|
|
Загрузка…
Ссылка в новой задаче