bump llama.cpp version to the latest

crates/llama_cpp/src/model/mod.rs

@@ -384,6 +384,7 @@ impl LlamaModel {
                 token.0,
                 buffer.as_mut_ptr() as *mut c_char,
                 std::os::raw::c_int::from(initial_size),
+                false,
             )
         };
 
@@ -398,6 +399,7 @@ impl LlamaModel {
                     token.0,
                     buffer.as_mut_ptr() as *mut c_char,
                     std::os::raw::c_int::from(buffer.len() as i32),
+                    false,
                 )
             };
             assert_eq!(size as usize, buffer.len(), "Buffer length doesn't match");
@@ -442,6 +444,7 @@ impl LlamaModel {
                     t.0,
                     token_buf.as_mut_ptr() as *mut c_char,
                     token_buf.len() as i32,
+                    false,
                 )
             };
 

crates/llama_cpp/src/model/params.rs

@@ -112,6 +112,7 @@ impl From<LlamaParams> for llama_model_params {
             vocab_only: value.vocab_only,
             use_mmap: value.use_mmap,
             use_mlock: value.use_mlock,
+            check_tensors: false,
         }
     }
 }

crates/llama_cpp/src/session/params.rs

@@ -349,6 +349,7 @@ impl From<SessionParams> for llama_context_params {
             pooling_type: value.pooling.into(),
             abort_callback: None,
             abort_callback_data: null_mut(),
+            flash_attn: false,
         }
     }
 }

crates/llama_cpp_sys/include/build-info.h

@@ -13,7 +13,7 @@
 #ifndef BUILD_INFO_H
 #define BUILD_INFO_H
 
-#define BUILD_NUMBER 2589
-#define BUILD_COMMIT "60cdf40"
+#define BUILD_NUMBER 3091
+#define BUILD_COMMIT "2b338967"
 
 #endif // BUILD_INFO_H