diff --git a/examples/llava/llava.cpp b/examples/llava/llava.cpp index 35049cc2c..48795f3f8 100644 --- a/examples/llava/llava.cpp +++ b/examples/llava/llava.cpp @@ -405,10 +405,8 @@ bool llava_validate_embed_size(const llama_context * ctx_llama, const clip_ctx * } bool llava_image_embed_make_with_clip_img(clip_ctx * ctx_clip, int n_threads, const clip_image_u8 * img, float ** image_embd_out, int * n_img_pos_out) { - int num_max_patches = 6; - if (clip_is_minicpmv(ctx_clip)) { - num_max_patches = 10; - } + // Minicpmv / granite vision use 10 patches + int num_max_patches = 10; if (clip_is_glm(ctx_clip)) { num_max_patches = 1; }