samuel ce447e2745 Add LLAMA_STATE_SEQ_FLAGS_APPEND for incremental KV state restore ... Add a new flag LLAMA_STATE_SEQ_FLAGS_APPEND (value 2) that skips the seq_rm() call in state_read_meta, allowing incremental chunk-by-chunk restore to the same sequence via repeated state_seq_set_data_ext calls. This enables external KV cache systems to restore opaque state blobs one chunk at a time without each chunk clearing the previous one. - Add #define LLAMA_STATE_SEQ_FLAGS_APPEND 2 in llama.h - Thread flags parameter through state_read() to state_read_meta() - Gate seq_rm() on !(flags & LLAMA_STATE_SEQ_FLAGS_APPEND) - Default behavior (flags=0) is unchanged		2026-03-31 10:46:18 +00:00
..
llama-cpp.h	llama : re-enable manual LoRA adapter free (#19983)	2026-03-18 12:03:26 +02:00
llama.h	Add LLAMA_STATE_SEQ_FLAGS_APPEND for incremental KV state restore	2026-03-31 10:46:18 +00:00