nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 20 Packages Projects Releases Wiki Activity

llama : document that only one on-device state can be saved per sequence (#23520)

Browse Source
This commit is contained in:
Tim Neumann
2026-05-25 10:29:28 +03:00
committed by GitHub
parent d55fb97174
commit 314e729347
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/llama.h
+2 -1
View File
@@ -874,7 +874,8 @@ extern "C" {
// work only with partial states, such as SWA KV cache or recurrent cache (e.g. Mamba) // work only with partial states, such as SWA KV cache or recurrent cache (e.g. Mamba)
#define LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY 1 #define LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY 1
// keeps the tensor data on device buffers (i.e. not accessible in host memory, but faster save/load) // Keeps the tensor data on device buffers (i.e. not accessible in host memory, but faster save/load).
// Getting the state for a seq_id with this flag invalidates all prior states gotten for that seq_id with this flag.
#define LLAMA_STATE_SEQ_FLAGS_ON_DEVICE 2 #define LLAMA_STATE_SEQ_FLAGS_ON_DEVICE 2
typedef uint32_t llama_state_seq_flags; typedef uint32_t llama_state_seq_flags;