0xarchit commited on
Commit
5118d6d
·
1 Parent(s): fd58bb8

bring back flash attention

Browse files
Files changed (1) hide show
  1. start.sh +3 -1
start.sh CHANGED
@@ -199,7 +199,9 @@ else
199
  server_args+=(--no-mmap)
200
  fi
201
 
202
-
 
 
203
 
204
  if [ "${NO_WARMUP}" = "1" ]; then
205
  server_args+=(--no-warmup)
 
199
  server_args+=(--no-mmap)
200
  fi
201
 
202
+ if [ "$FLASH_ATTN" = "1" ]; then
203
+ server_args+=(--flash-attn on)
204
+ fi
205
 
206
  if [ "${NO_WARMUP}" = "1" ]; then
207
  server_args+=(--no-warmup)