From 29b6e9a5c832b7d148044b45fb3cd6f3f923d96b Mon Sep 17 00:00:00 2001
From: Andrei Betlen <abetlen@gmail.com>
Date: Tue, 30 Apr 2024 09:32:47 -0400
Subject: [PATCH] fix: wrong parameter for flash attention in pickle
 __getstate__

---
 llama_cpp/llama.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llama_cpp/llama.py b/llama_cpp/llama.py
index 172f4c6..f927f0c 100644
--- a/llama_cpp/llama.py
+++ b/llama_cpp/llama.py
@@ -1777,7 +1777,7 @@ class Llama:
             logits_all=self.context_params.logits_all,
             embedding=self.context_params.embeddings,
             offload_kqv=self.context_params.offload_kqv,
-            flash_offload=self.context_params.flash_offload,
+            flash_attn=self.context_params.flash_attn,
             # Sampling Params
             last_n_tokens_size=self.last_n_tokens_size,
             # LoRA Params