mtmd : add not supported msg for qwen2vl

ngxson · ngxson · commit 7360d43a7655 · 2025-04-19T09:08:46.000+02:00
diff --git a/examples/llava/mtmd.cpp b/examples/llava/mtmd.cpp
@@ -58,6 +58,8 @@ struct mtmd_context {
         }
         this->text_model = text_model;
 
+        GGML_ASSERT(!clip_is_qwen2vl(ctx_clip) && "Qwen2VL model is not supported yet, use llama-qwen2vl-cli instead");
+
         int minicpmv_version = clip_is_minicpmv(ctx_clip);
         if (minicpmv_version == 2) {
             // minicpmv 2.5 format:
@@ -504,6 +506,7 @@ int32_t mtmd_helper_eval(mtmd_context * ctx,
 
             if (mtmd_decode_use_non_causal(ctx)) {
                 llama_set_causal_attn(lctx, false);
+                // TODO @ngxson : need to make sure only one image is processed at a time, and n_ubatch must be enough to hold the image
             }
 
             while (i_batch < n_img_batches) { // split into batches

Original file line number	Diff line number	Diff line change
`@@ -58,6 +58,8 @@ struct mtmd_context {`
`58`	`58`	`}`
`59`	`59`	`this->text_model = text_model;`
`60`	`60`
	`61`	`+ GGML_ASSERT(!clip_is_qwen2vl(ctx_clip) && "Qwen2VL model is not supported yet, use llama-qwen2vl-cli instead");`
	`62`	`+`
`61`	`63`	`int minicpmv_version = clip_is_minicpmv(ctx_clip);`
`62`	`64`	`if (minicpmv_version == 2) {`
`63`	`65`	`// minicpmv 2.5 format:`
`@@ -504,6 +506,7 @@ int32_t mtmd_helper_eval(mtmd_context * ctx,`
`504`	`506`
`505`	`507`	`if (mtmd_decode_use_non_causal(ctx)) {`
`506`	`508`	`llama_set_causal_attn(lctx, false);`
	`509`	`+ // TODO @ngxson : need to make sure only one image is processed at a time, and n_ubatch must be enough to hold the image`
`507`	`510`	`}`
`508`	`511`
`509`	`512`	`while (i_batch < n_img_batches) { // split into batches`