Set kv cache size to f16 by default

2023-04-14 22:21:19 -04:00 · 2023-04-14 22:21:19 -04:00 · 6e298d8fca
parent 9c8c2c37dc
commit 6e298d8fca
1 changed files with 1 additions and 1 deletions
--- a/llama_cpp/llama.py
+++ b/llama_cpp/llama.py
@ -21,7 +21,7 @@ class Llama:
        n_ctx: int = 512,
        n_parts: int = -1,
        seed: int = 1337,
-        f16_kv: bool = False,
+        f16_kv: bool = True,
        logits_all: bool = False,
        vocab_only: bool = False,
        use_mmap: bool = True,