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 5, 12:15 UTC
[EASY] Drop duplicate KV-cache initialization (#38799)
Signed-off-by: namgyu-youn <namgyu.dev@gmail.com>
namgyu-youn committed
Apr 7, 2026 at 03:05 UTC
94fbb09894a00533a41ce2d976d9aa2f06e7e000
1 file changed
-3
vllm/model_executor/layers/attention/attention.py
-3
@@ -131,9 +131,6 @@ def _init_kv_cache_quant(
131
quant_config: Optional quantization configuration.
132
prefix: Layer name prefix for quantization method lookup.
133
"""
134
- quant_method = (
135
- quant_config.get_quant_method(layer, prefix=prefix) if quant_config else None
136
- )
134
135
# Note [Register q/k/v/prob scales in state dict]
136
# When calling model.to(device), only parameters/buffers in state dict are