still does not work?

Author: ngxson

tools/llava/clip.cpp

@@ -1204,18 +1204,14 @@ static ggml_cgraph * clip_image_build_graph_ultravox(clip_ctx * ctx, const clip_
         // mid-norm
         cur = ggml_rms_norm(ctx0, cur, 1e-6);
         cur = ggml_mul(ctx0, cur, model.mm_norm_mid_w);
+        embeddings = cur;
 
         // ffn out
         cur = ggml_mul_mat(ctx0, model.mm_2_w, cur);
 
         embeddings = cur;
     }
 
-    embeddings = ggml_view_2d(ctx0, embeddings, 2048, n_step / 16,
-                              ggml_row_size(embeddings->type, 2048), 0);
-
-    printf("shape of embd: %lld %lld %lld\n", embeddings->ne[0], embeddings->ne[1], embeddings->ne[2]);
-
     // build the graph
     ggml_build_forward_expand(gf, embeddings);
 

tools/llava/mtmd.cpp

@@ -144,13 +144,13 @@ mtmd_context * mtmd_init_from_file(const char * mmproj_fname,
         clip_image_f32_batch * batch = clip_image_f32_batch_init();
         std::vector<float> mel(128 * 1024);
         clip_image_f32_batch_add_mel(batch, 128, 1024, mel.data());
-        std::vector<float> output(64 * 2048);
+        std::vector<float> output(64 * 2048, 0);
         clip_image_batch_encode(test->ctx_clip, 8, batch, output.data());
         for (int i = 0; i < 3; i++) printf("%f ", output[i]); printf("\n");
         for (int i = 0; i < 3; i++) printf("%f ", output[i+2048]); printf("\n");
         for (int i = 0; i < 3; i++) printf("%f ", output[i+2048*2]); printf("\n");
         float sum = 0.0;
-        for (size_t i = 0; i < 1000; i++) sum += output[i];
+        for (size_t i = 0; i < 64 * 2048; i++) sum += output[i];
         printf("sum: %f\n", sum);
         GGML_ABORT("test");
     } catch (const std::exception & e) {