diff --git a/llama.cpp b/llama.cpp
index a4cb1f284..d9e87f34f 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -13233,7 +13233,7 @@ void llama_sample_min_p(struct llama_context * ctx, llama_token_data_array * can
     }
 }
 
-void llama_sample_dry(llama_token_data_array * candidates, const llama_token * last_tokens, int last_tokens_size, float dry_base, float dry_multiplier, int dry_allowed_length, const llama_token * dry_seq_breakers, int dry_seq_breakers_size) {
+void llama_sample_dry(llama_token_data_array * candidates, const llama_token * last_tokens, size_t last_tokens_size, float dry_base, float dry_multiplier, int dry_allowed_length, const llama_token * dry_seq_breakers, size_t dry_seq_breakers_size) {
     // skip dry sampler if we don't have a previous token
     if (last_tokens_size < 1) return;
 
diff --git a/llama.h b/llama.h
index 774c1b222..fbba5daf8 100644
--- a/llama.h
+++ b/llama.h
@@ -928,12 +928,12 @@ extern "C" {
     LLAMA_API void llama_sample_dry(
           llama_token_data_array * candidates,
                const llama_token * last_tokens,
-                             int   last_tokens_size,
+                          size_t   last_tokens_size,
                            float   dry_base,
                            float   dry_multiplier,
                              int   dry_allowed_length,
                const llama_token * dry_seq_breakers,
-                             int   dry_seq_breakers_size);
+                          size_t   dry_seq_breakers_size);
 
     /// @details Tail Free Sampling described in https://www.trentonbricken.com/Tail-Free-Sampling/.
     LLAMA_API void llama_sample_tail_free(