Enable overlap by default (#2067)

This commit is contained in:
Lianmin Zheng
2024-11-19 22:07:58 -08:00
committed by GitHub
parent 699384cb01
commit 7d671e4ad2
17 changed files with 92 additions and 75 deletions

View File

@@ -899,10 +899,7 @@ class ScheduleBatch:
self.input_ids = self.output_ids
self.output_ids = None
if self.sampling_info.penalizer_orchestrator:
self.sampling_info.penalizer_orchestrator.cumulate_output_tokens(
self.input_ids
)
self.sampling_info.penalizer_orchestrator.cumulate_output_tokens(self.input_ids)
# Alloc mem
bs = len(self.reqs)