llama : bump max seq limit from 64 to 256 (#15916)

ggerganov · web-flow · commit e58174cecbc4 · 2025-09-18T12:47:56.000+03:00
ggml-ci
diff --git a/src/llama-cparams.h b/src/llama-cparams.h
@@ -4,7 +4,7 @@
 
 #include <cstdint>
 
-#define LLAMA_MAX_SEQ 64
+#define LLAMA_MAX_SEQ 256
 
 struct llama_cparams {
     uint32_t n_ctx;           // context size used during inference
