mtmd : DeepSeek-OCR image processing fixes, img_tool::resize padding refactor (ggml-org#23345)

* mtmd : deepseek-ocr fixes, improvements and refactoring

- image processing changes to achieve full parity with Pillow (reference impl)
- SAM mask casting only when flash-attn is on
- SAM refactor (build_sam() extracted so deepseek-ocr-2 can reuse it)
- llama-chat changes to fix server/WebUI issue (new media_markers_first())
- adapted test-chat-template and added test cases for deepseek-ocr
- changed regression test for deepseek-ocr to use CER+chrF scores for ground-truth comparison; removed embedding-model
- ty.toml ignore unresolved-import for tools/mtmd/tests/**

* image-text reordering fix removed

* refactor bool add_padding + pad_rounding enum into a single pad_style enum

Author: sfallah

tools/mtmd/clip-model.h

@@ -35,6 +35,16 @@ enum resize_algo {
     // RESIZE_ALGO_LANCZOS, // TODO
 };
 
+// Padding style for img_tool::resize
+//   PAD_NONE    - no padding; direct resize to target dimensions
+//   PAD_CEIL    - aspect-preserving pad (default)
+//   PAD_NEAREST - aspect-preserving pad with nearest-integer rounding (Pillow byte-parity)
+enum pad_style {
+    PAD_NONE,
+    PAD_CEIL,
+    PAD_NEAREST,
+};
+
 struct clip_hparams {
     int32_t image_size = 0;
     int32_t patch_size = 0;
@@ -52,7 +62,7 @@ struct clip_hparams {
     int32_t image_min_pixels = -1;
     int32_t image_max_pixels = -1;
     resize_algo image_resize_algo = RESIZE_ALGO_BICUBIC;
-    bool image_resize_pad = true; // if false, center-crop will be applied when resizing
+    pad_style image_resize_pad = PAD_CEIL; // padding style when resizing
     std::array<uint8_t, 3> image_pad_color = {0, 0, 0};
 
     // (preprocessor) for llava-uhd style models
@@ -61,8 +71,8 @@ struct clip_hparams {
     int32_t preproc_max_tiles = 0;
     resize_algo image_resize_algo_rf = RESIZE_ALGO_BICUBIC;
     resize_algo image_resize_algo_ov = RESIZE_ALGO_BILINEAR;
-    bool image_pad_rf = true;  // if true, refined image will be padded (e.g. llava-1.6)
-    bool image_pad_ov = false; // if true, overview image will be padded (e.g. llava-1.6)
+    pad_style image_pad_rf = PAD_CEIL;  // padding style for the refined image (e.g. llava-1.6)
+    pad_style image_pad_ov = PAD_NONE;  // padding style for the overview image (e.g. llava-1.6)
     std::array<uint8_t, 3> image_pad_color_rf = {0, 0, 0}; // padding color for refined image
     std::array<uint8_t, 3> image_pad_color_ov = {0, 0, 0}; // padding color for overview image
 

tools/mtmd/clip.cpp

@@ -1233,20 +1233,20 @@ struct clip_model_loader {
                         hparams.has_llava_projector = model.proj_type != PROJECTOR_TYPE_COGVLM;
                         hparams.image_pad_color     = {122, 116, 104};
                         if (!hparams.image_res_candidates.empty()) {
-                            hparams.image_resize_pad  = true;
+                            hparams.image_resize_pad  = PAD_CEIL;
                             hparams.image_resize_algo = RESIZE_ALGO_BILINEAR;
                         } else {
                             // llava-1.6 default params
-                            hparams.image_pad_ov         = false;
-                            hparams.image_pad_rf         = true;
+                            hparams.image_pad_ov         = PAD_NONE;
+                            hparams.image_pad_rf         = PAD_CEIL;
                             hparams.image_pad_color_rf   = {122, 116, 104};
                             hparams.image_resize_algo_rf = RESIZE_ALGO_BICUBIC;
                             hparams.image_resize_algo_ov = RESIZE_ALGO_BILINEAR;
                         }
                     } break;
                 case PROJECTOR_TYPE_GLM_EDGE:
                     {
-                        hparams.image_resize_pad  = true;
+                        hparams.image_resize_pad  = PAD_CEIL;
                         hparams.image_resize_algo = RESIZE_ALGO_BILINEAR;
                     } break;
                 case PROJECTOR_TYPE_MINICPMV:
@@ -1441,7 +1441,7 @@ struct clip_model_loader {
                     {
                         hparams.n_merge = 2;
                         hparams.image_resize_algo = RESIZE_ALGO_BILINEAR;
-                        hparams.image_resize_pad  = false;
+                        hparams.image_resize_pad  = PAD_NONE;
                         get_u32(KEY_SPATIAL_MERGE_SIZE, hparams.n_merge, false);
                         get_u32(KEY_ATTN_WINDOW_SIZE, hparams.attn_window_size, true);
                         std::vector<int> wa_layer_indexes_vec;
@@ -1461,7 +1461,7 @@ struct clip_model_loader {
 
                         // reka model performs better when using resize_bicubic, which stretches
                         // the image to fit fixed square size
-                        hparams.image_resize_pad = false;
+                        hparams.image_resize_pad = PAD_NONE;
                     } break;
                 case PROJECTOR_TYPE_GLM4V:
                     {
@@ -1516,9 +1516,7 @@ struct clip_model_loader {
                         hparams.image_size = 1024;
                         hparams.warmup_image_size = 1024;
                         hparams.image_resize_algo = RESIZE_ALGO_BICUBIC_PILLOW;
-                        hparams.image_pad_color[0] = hparams.image_mean[0];
-                        hparams.image_pad_color[1] = hparams.image_mean[1];
-                        hparams.image_pad_color[2] = hparams.image_mean[2];
+                        hparams.image_pad_color = {127, 127, 127};
 
                         get_u32(KEY_SAM_N_BLOCK, hparams.sam_n_layer, true);
                         get_u32(KEY_SAM_N_HEAD, hparams.sam_n_head, true);
@@ -1537,7 +1535,7 @@ struct clip_model_loader {
                     {
                         hparams.n_merge = 2;
                         hparams.image_resize_algo = RESIZE_ALGO_BICUBIC_PILLOW;
-                        hparams.image_resize_pad = false;
+                        hparams.image_resize_pad = PAD_NONE;
                         hparams.ffn_op = FFN_GELU;
                         get_u32(KEY_SPATIAL_MERGE_SIZE, hparams.n_merge, false);
                         hparams.set_limit_image_tokens(256, 16384);