apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature][MTP]Support MTP for rl-model (#4009)

Browse Source 
* qk norm for speculate decode C16

* support mtp in v1_scheduler mode

* support mtp rope_3d

* support mtp features

* add unit test && del some log

---------

Co-authored-by: yuanxiaolan <yuanxiaolan01@baidu.com>
Co-authored-by: xiaoxiaohehe001 <hiteezsf@163.com>
This commit is contained in:
freeliuzc
2025-09-10 13:34:37 +08:00
committed by GitHub
parent cce2410fad
commit 2f473ba966
21 changed files with 1465 additions and 531 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
fastdeploy/output/token_processor.py
View File

@@ -261,7 +261,7 @@ class TokenProcessor:
def _compute_speculative_status(self):
# TODO(liuzichang): Supplement more statistics
interval = 10
interval = 1
if self.speculative_stats_step % interval == 0:
accept_ratio = 1 - self.total_step * 1.0 / self.number_of_output_tokens
spec_logger.info(