lhutyra
diff --git a/‎llama_cpp/llama.py
Copy file name to clipboardExpand all lines: llama_cpp/llama.py
+1-1Lines changed: 1 addition & 1 deletion b/‎llama_cpp/llama.py
Copy file name to clipboardExpand all lines: llama_cpp/llama.py
+1-1Lines changed: 1 addition & 1 deletion
diff --git a/‎llama_cpp/server/settings.py
Copy file name to clipboardExpand all lines: llama_cpp/server/settings.py
+1-1Lines changed: 1 addition & 1 deletion b/‎llama_cpp/server/settings.py
Copy file name to clipboardExpand all lines: llama_cpp/server/settings.py
+1-1Lines changed: 1 addition & 1 deletion
@@ -77,7 +77,7 @@ def __init__(
         mul_mat_q: bool = True,
         logits_all: bool = False,
         embedding: bool = False,
-        offload_kqv: bool = False,
+        offload_kqv: bool = True,
         # Sampling Params
         last_n_tokens_size: int = 64,
         # LoRA Params
 
@@ -90,7 +90,7 @@ class ModelSettings(BaseSettings):
     logits_all: bool = Field(default=True, description="Whether to return logits.")
     embedding: bool = Field(default=True, description="Whether to use embeddings.")
     offload_kqv: bool = Field(
-        default=False, description="Whether to offload kqv to the GPU."
+        default=True, description="Whether to offload kqv to the GPU."
     )
     # Sampling Params
     last_n_tokens_size: int = Field(