apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-09-26 20:41:53 +08:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

[Optimize]support machete weight only gemm (#3561)

Browse Source 
* support machete weight only gemm

* add generate

* update

* fix

* change file location

* add sm_version limit

* fix

* fix

* fix ci

* fix coverage

* fix xpu
This commit is contained in:
Sunny-bot1
2025-08-28 09:49:58 +08:00
committed by GitHub
parent e37e86b3b8
commit 479c8b85d3
29 changed files with 5436 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
fastdeploy/envs.py
View File

@@ -54,6 +54,8 @@ environment_variables: dict[str, Callable[[], Any]] = {
"FD_SAMPLING_CLASS": lambda: os.getenv("FD_SAMPLING_CLASS", "base"),
# Set moe backend."cutlass","marlin" and "triton" can be set currently.
"FD_MOE_BACKEND": lambda: os.getenv("FD_MOE_BACKEND", "cutlass"),
# Whether to use Machete for wint4 dense gemm.
"FD_USE_MACHETE": lambda: os.getenv("FD_USE_MACHETE", "0"),
# Set whether to disable recompute the request when the KV cache is full.
"FD_DISABLED_RECOVER": lambda: os.getenv("FD_DISABLED_RECOVER", "0"),
# Set triton kernel JIT compilation directory.