| .. |
|
blocksparse_attention
|
7d79c0e726
chore: use nvml query to avoid accidental cuda initialization
|
7 ay önce |
|
__init__.py
|
9d81716bfd
[v0.5.3] Release Candidate (#388)
|
10 ay önce |
|
ipex_attn.py
|
805fa8721d
feat: use intel_extension_for_pytorch for CPU backend
|
7 ay önce |
|
paged_attn.py
|
156f577f79
feat: switch from `PYBIND11_MODULE` to `TORCH_LIBRARY` (#569)
|
7 ay önce |
|
prefix_prefill.py
|
7d79c0e726
chore: use nvml query to avoid accidental cuda initialization
|
7 ay önce |
|
triton_flash_attn.py
|
8e11259e90
missing triton autoconfig for rocm flash attn
|
8 ay önce |
AlpinDale
7d79c0e726
chore: use nvml query to avoid accidental cuda initialization