apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature][MTP]support new speculative decoding method named hybrid mtp with ngram (#3610)

Browse Source
This commit is contained in:
freeliuzc
2025-08-26 14:29:22 +08:00
committed by GitHub
parent 0a0d2959b9
commit 52eda7fdb3
20 changed files with 454 additions and 571 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
fastdeploy/model_executor/pre_and_post_process.py
View File

@@ -68,7 +68,7 @@ else:
speculate_set_value_by_flags_and_idx,
speculate_step_paddle,
speculate_step_system_cache,
speculate_update_v3,
speculate_update,
step_paddle,
step_system_cache,
update_inputs,
@@ -308,7 +308,7 @@ def post_process_normal(
def post_process_specualate(model_output, save_each_rank: bool = False, skip_save_output: bool = False):
""""""
speculate_update_v3(
speculate_update(
model_output.seq_lens_encoder,
model_output.seq_lens_decoder,
model_output.not_need_stop,