apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-18 22:44:39 +08:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

[BugFix]Fix sample rejection (#2908)

Browse Source 
* fix config

* fix rejection
This commit is contained in:
YuanRisheng
2025-07-18 13:44:30 +08:00
committed by GitHub
parent e679567d59
commit 0eb5dc18d3
2 changed files with 1 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
fastdeploy/config.py
View File

@@ -70,7 +70,7 @@ class ModelConfig:
self.stop_seqs_max_len = 8 self.stop_seqs_max_len = 8
# NOTE(gongshaotain): form _load_model_init_val() # NOTE(gongshaotain): form _load_model_init_val()
self.top_p = 0.0 self.top_p = 1.0
self.temperature = 1.0 self.temperature = 1.0
self.rope_theta = 10000.0 self.rope_theta = 10000.0
self.penalty_score = 1.0 self.penalty_score = 1.0

1
fastdeploy/worker/gpu_model_runner.py
View File

@@ -419,7 +419,6 @@ class GPUModelRunner(ModelRunnerBase):
self.share_inputs["max_dec_len"][idx:idx + 1] = max_dec_len self.share_inputs["max_dec_len"][idx:idx + 1] = max_dec_len
self.share_inputs["min_dec_len"][idx:idx + 1] = max_dec_len self.share_inputs["min_dec_len"][idx:idx + 1] = max_dec_len
self.share_inputs["stop_flags"][idx:idx + 1] = False self.share_inputs["stop_flags"][idx:idx + 1] = False
self.share_inputs["top_p"][idx:idx + 1] = 0.0
self.share_inputs["temperature"][idx:idx + 1] = 1 self.share_inputs["temperature"][idx:idx + 1] = 1
self.share_inputs["first_token_ids"][ self.share_inputs["first_token_ids"][