| .. |
|
all_reduce
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
attention
|
f831fd8312
rocm: fix compile issues with rocm 6.2 (#893)
|
3 周之前 |
|
backup
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
9 月之前 |
|
core
|
0256ed236b
feat: windows support (#790)
|
2 月之前 |
|
cpu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
cutlass_extensions
|
93bc863591
feat: Machete Kernels for Hopper GPUs (#842)
|
1 月之前 |
|
hadamard
|
5d288aa76c
feat: add fast hadamard transformation kernels (#232)
|
11 月之前 |
|
mamba
|
0256ed236b
feat: windows support (#790)
|
2 月之前 |
|
moe
|
0256ed236b
feat: windows support (#790)
|
2 月之前 |
|
prepare_inputs
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
punica
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
quantization
|
e14223dce5
kernel: use `cub::BlockReduce` instead of custom impl (#895)
|
3 周之前 |
|
sampling
|
bfc8988116
feat: add cuda sampling kernels for top_k and top_p (#828)
|
1 月之前 |
|
activation_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
cache.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
cache_kernels.cu
|
0256ed236b
feat: windows support (#790)
|
2 月之前 |
|
cuda_compat.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
cuda_utils.h
|
93bc863591
feat: Machete Kernels for Hopper GPUs (#842)
|
1 月之前 |
|
cuda_utils_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
dispatch_utils.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
layernorm_kernels.cu
|
e14223dce5
kernel: use `cub::BlockReduce` instead of custom impl (#895)
|
3 周之前 |
|
ops.h
|
93bc863591
feat: Machete Kernels for Hopper GPUs (#842)
|
1 月之前 |
|
permute_cols.cu
|
93bc863591
feat: Machete Kernels for Hopper GPUs (#842)
|
1 月之前 |
|
pos_encoding_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 月之前 |
|
torch_bindings.cpp
|
93bc863591
feat: Machete Kernels for Hopper GPUs (#842)
|
1 月之前 |
AlpinDale
e14223dce5
kernel: use `cub::BlockReduce` instead of custom impl (#895)