apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[BugFix] fix fused_mask_swiglu_fp8_quant bug (#6316)

Browse Source 
* optimize mask_quant op speed up 1.5

* fix calculate sequence

* add fused

* rm log

* push kernel code

* add ut

* accuracy ok

* add ue8m0

* add ut

* add merge develop

* rm ut of mask_per_token_quant

* Revert "[Optimize] optimize mask_quant & swiglu (#6222)"

This reverts commit 2ada119a38.

* add block_size

* pre-commit
This commit is contained in:
fxyfxy777
2026-02-03 13:54:12 +08:00
committed by GitHub
parent 6563b8307c
commit f3413c4caa
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/layers/moe/fused_moe_deepgemm_backend.py
+1 -1
View File
@@ -414,7 +414,7 @@ class DeepGemmFusedMoeMethod(MoEMethodBase):
)
act_out_fp8, scale = fastdeploy.model_executor.ops.gpu.fused_mask_swiglu_fp8_quant(
up_gate_proj_out, token_nums_per_expert, use_ue8m0=False
up_gate_proj_out, token_nums_per_expert, block_size=128, use_ue8m0=False
)
deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_masked(