| .. |
|
all_reduce
|
29eaded422
fix and re-enable custom all-reduce
|
9 сар өмнө |
|
attention
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
backup
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
9 сар өмнө |
|
cpu
|
fb23720c72
fix CPU build
|
9 сар өмнө |
|
hadamard
|
5d288aa76c
feat: add fast hadamard transformation kernels (#232)
|
11 сар өмнө |
|
moe
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
punica
|
c0aac15421
feat: S-LoRA support (#222)
|
1 жил өмнө |
|
quantization
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
activation_kernels.cu
|
3d6695cfbb
feat: add approximate gelu activation kernels (#370)
|
9 сар өмнө |
|
cache.h
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
cache_kernels.cu
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
cuda_compat.h
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
cuda_utils.h
|
31c95011a6
feat: FP8 E5M2 KV Cache (#226)
|
1 жил өмнө |
|
cuda_utils_kernels.cu
|
31c95011a6
feat: FP8 E5M2 KV Cache (#226)
|
1 жил өмнө |
|
dispatch_utils.h
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
9 сар өмнө |
|
layernorm_kernels.cu
|
40f63268ee
disable new layernorm kernels for CUDA < 12.0
|
9 сар өмнө |
|
ops.h
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
pos_encoding_kernels.cu
|
e702f587cf
feat: add batched RoPE kernels (#371)
|
9 сар өмнө |
|
pybind.cpp
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
|
reduction.cuh
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 сар өмнө |
AlpinDale
40f63268ee
disable new layernorm kernels for CUDA < 12.0