AlpinDale ad1c6b86a1 gptq_marlin: enable bfloat16 7 mesiacov pred
..
all_reduce 9d81716bfd [v0.5.3] Release Candidate (#388) 10 mesiacov pred
attention 251568470e initial nvidia fp8 e4m3 for kv cache 7 mesiacov pred
backup f8dfac6372 chore: attention refactor and upstream sync apr01 (#365) 11 mesiacov pred
cpu b746fb5562 fix a few warnings on the cpu kernels 7 mesiacov pred
hadamard 5d288aa76c feat: add fast hadamard transformation kernels (#232) 1 rok pred
moe 9d81716bfd [v0.5.3] Release Candidate (#388) 10 mesiacov pred
punica e3f2ea4850 make punica kernels work with rocm 7 mesiacov pred
quantization ad1c6b86a1 gptq_marlin: enable bfloat16 7 mesiacov pred
activation_kernels.cu 3d6695cfbb feat: add approximate gelu activation kernels (#370) 11 mesiacov pred
cache.h 251568470e initial nvidia fp8 e4m3 for kv cache 7 mesiacov pred
cache_kernels.cu 251568470e initial nvidia fp8 e4m3 for kv cache 7 mesiacov pred
cuda_compat.h e3f2ea4850 make punica kernels work with rocm 7 mesiacov pred
cuda_utils.h 31c95011a6 feat: FP8 E5M2 KV Cache (#226) 1 rok pred
cuda_utils_kernels.cu 31c95011a6 feat: FP8 E5M2 KV Cache (#226) 1 rok pred
dispatch_utils.h f8dfac6372 chore: attention refactor and upstream sync apr01 (#365) 11 mesiacov pred
layernorm_kernels.cu 9d81716bfd [v0.5.3] Release Candidate (#388) 10 mesiacov pred
ops.h 35ae01d7ba refactor: attention metadata term 7 mesiacov pred
pos_encoding_kernels.cu e702f587cf feat: add batched RoPE kernels (#371) 11 mesiacov pred
pybind.cpp 2351a0e2cd feat: FlashInfer backend for decoding phase (#548) 7 mesiacov pred
reduction.cuh 9d81716bfd [v0.5.3] Release Candidate (#388) 10 mesiacov pred