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

fix mtp acceptance rate decline (#6470)

Browse Source
This commit is contained in:
kevin
2026-02-12 19:56:10 +08:00
committed by GitHub
parent 51f812aaa4
commit 52edf5e9b3
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/worker/input_batch.py
+2 -1
View File
@@ -890,7 +890,8 @@ class ProposerInputBatch(InputBatch):
self.block_tables = paddle.clone(self.target_model_input_batch["block_tables"])
self.input_ids = paddle.clone(self.target_model_input_batch["input_ids"])
fill_paddle_tensor(self, "input_ids_cpu", -1)
self.seq_lens_this_time_buffer = paddle.clone(self.target_model_input_batch["seq_lens_this_time"])
# acceptance rate decline when reset seq_lens_this_time
# self.seq_lens_this_time_buffer = paddle.clone(self.target_model_input_batch["seq_lens_this_time"])
self.seq_lens_encoder = paddle.clone(self.target_model_input_batch["seq_lens_encoder"])
self.seq_lens_decoder = paddle.clone(self.target_model_input_batch["seq_lens_decoder"])