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

[BugFix] Add prefill restrictions for chunked_prefill+VL (#2983)

Browse Source
This commit is contained in:
Zero Rains
2025-07-23 16:45:57 +08:00
committed by GitHub
parent a39a67334c
commit 850c9d98d4
2 changed files with 18 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
fastdeploy/worker/gpu_model_runner.py
View File

@@ -150,10 +150,17 @@ class GPUModelRunner(ModelRunnerBase):
"""
Check whether prefill stage finished
"""
if int(paddle.max(self.share_inputs["seq_lens_encoder"])) != 0:
return 1
if self.enable_mm:
# VL only support 1 batch to prefill
prefill_statue = (self.share_inputs["seq_lens_this_time"] != 0) & (
self.share_inputs["seq_lens_this_time"] != 1
)
return not paddle.any(prefill_statue).numpy()
else:
return 0
if int(paddle.max(self.share_inputs["seq_lens_encoder"])) != 0:
return 1
else:
return 0
def _init_speculative_proposer(self):
"""