.. |
all_reduce
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
attention
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
backup
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
9 mēneši atpakaļ |
core
|
6144150398
chore: use scalar type to dispatch to different `gptq_marlin` kernels (#689)
|
4 mēneši atpakaļ |
cpu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
hadamard
|
5d288aa76c
feat: add fast hadamard transformation kernels (#232)
|
11 mēneši atpakaļ |
mamba
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
moe
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
prepare_inputs
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
punica
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
quantization
|
6144150398
chore: use scalar type to dispatch to different `gptq_marlin` kernels (#689)
|
4 mēneši atpakaļ |
activation_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
cache.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
cache_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
cuda_compat.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
cuda_utils.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
cuda_utils_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
dispatch_utils.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
layernorm_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
ops.h
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
pos_encoding_kernels.cu
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
reduction.cuh
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mēneši atpakaļ |
torch_bindings.cpp
|
a401f8e05d
feat: per-tensor token epilogue kernels (#630)
|
4 mēneši atpakaļ |