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