use a bool mask for attention (#1319)

2025-12-16 02:08:55 +08:00 · 2025-03-04 12:47:32 -08:00
parent 1bc3476a46
commit 65aa2ec849
2 changed files with 5 additions and 6 deletions
--- a/llms/mlx_lm/generate.py
+++ b/llms/mlx_lm/generate.py
@@ -152,7 +152,7 @@ def setup_arg_parser():
        "--num-draft-tokens",
        type=int,
        help="Number of tokens to draft when using speculative decoding.",
-        default=2,
+        default=3,
    )
    return parser