apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Quantization] Support w4afp8 MoE dynamic quantization (#5282)

Browse Source 
* support dynamic activation quant for w4afp8

* support dynamic w4afp8

* add test

* fix

* fix

---------

Co-authored-by: zhoutianzi666 <17801055074@163.com>
This commit is contained in:
Sunny-bot1
2025-12-02 18:56:16 +08:00
committed by GitHub
parent 429dd2b1db
commit 3629db4129
9 changed files with 478 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
fastdeploy/model_executor/layers/quantization/__init__.py
View File

@@ -84,6 +84,13 @@ def parse_quant_config(args, model_config, is_ernie, is_v1_loader):
quantization_config["moe_quant_type"] = "wint4"
quantization_config["quantization"] = "mix_quant"
quant_config_name = "mix_quant"
# Special handling for moe w4afp8 dynamic quant
elif quant_config_name == "w4afp8":
quantization_config["dense_quant_type"] = "block_wise_fp8"
quantization_config["moe_quant_type"] = "w4afp8"
quantization_config["hadamard_block_size"] = 512
quantization_config["quantization"] = "mix_quant"
quant_config_name = "mix_quant"
else:
quant_config_name = None
if quant_config_name is None: