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 16, 23:15 UTC
Correct alignment in the seq_len diagram. (#5592)
Co-authored-by: Liqian Chen <liqian.chen@deeplang.ai>
Charles Riggins committed
Jun 18, 2024 at 00:05 UTC
9e74d9d003d546c17dca472c3f4b48be651f1d7c
1 file changed
+1
-1
vllm/attention/backends/flash_attn.py
+1
-1
@@ -83,7 +83,7 @@ class FlashAttentionMetadata(AttentionMetadata):
83
# |---------------- N iteration ---------------------|
84
# |- tokenA -|......................|-- newTokens ---|
85
# |---------- context_len ----------|
86
- # |-------------------- seq_len ----------------------|
86
+ # |-------------------- seq_len ---------------------|
87
# |-- query_len ---|
88
89
# Maximum query length in the batch. None for decoding.