A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 17, 01:45 UTC
[Misc] remove chunk detected debug logs (#4571)
DefTruth committed
May 3, 2024 at 12:48 UTC
ce3f1eedf8e7e015054a166f17205eb3206e4625
1 file changed
+4
-4
vllm/engine/llm_engine.py
+4
-4
@@ -662,10 +662,10 @@ class LLMEngine:
662
# decode seq_groups in scheduled_seq_groups.
663
if scheduler_outputs is not None:
664
num_generation_tokens_from_prefill_groups = 0.
665
- if scheduler_outputs.num_prefill_groups > 0 and len(
666
- scheduler_outputs.scheduled_seq_groups
667
- ) != scheduler_outputs.num_prefill_groups:
668
- print("DETECTED CHUNKED")
665
+ # NOTE: if scheduler_outputs.num_prefill_groups > 0 and
666
+ # the len of scheduler_outputs.scheduled_seq_groups is !=
667
+ # scheduler_outputs.num_prefill_groups, this means that
668
+ # chunked prefills have been detected.
669
670
for idx, scheduled_seq_group in enumerate(
671
scheduler_outputs.scheduled_seq_groups):