mtmd: Add Gemma3n multimodal support with MobileNetV5 vision encoder (#18256)

* Add Gemma3nVisionModel - MobileNetV5 vision encoder convertor to convert_hf_to_gguf.py. Add gemma3n to vision projectors in gguf-py/gguf/constants.py.

* Add mobilenetv5 impl

* Fix comments, remove unused vars

* Fix permute and remove transpose of projection weights

* Fix comments, remove debugging prints from hf_to_gguf

* 1. Hard-code image_mean = 0 and image_std = 1
2. Use available tensor mapping logic
3. Remove redundant chat template replacement of soft tokens placeholder with media placeholder

* 1. Move mobilenetv5 helpers declarations to `clip_graph_mobilenetv5` struct and definitions to mobilenetv5.cpp
2.Remove unused `clip_is_gemma3n` func declarations and definitions
3. Remove redundant `rescale_image_u8_to_f32` func and use `normalize_image_u8_to_f32` with zero mean and unit std
4. Calculate n_patches using image_size / patch_size

* Remove obsolete comments

* - convert_hf_to_gguf.py & constants.py & tensor_mapping.py: Use explicit mapping: Custom map for double indexed blocks and tensor_mapping.py for rest
- convert_hf_to_gguf.py: Unsqueeze Stem Bias and Layer scale tensors to correct shape while converting to gguf
- mobilenetv5.cpp: Remove explicit reshaping of Stem Bias and Layer scale which are now handled while converting to gguf, replace fprintf with LOG_*
- clip.cpp: Remove unused embedding and hard_emb_norm tensor loading

* - Rename tensors to v.conv..., v.blk..., v.msfa... to better align with already existing terminology

* Fix stem conv bias name

* Remove explicit handling of bias term for stem conv

* - Change order of addition in "project_per_layer_inputs" to support broadcasting of vision inp_per_layer
- Simplify the vision embeddings path of "get_per_layer_inputs" to output [n_embd_altup, n_layer, 1], broadcastable

* clean up conversion script

* fix code style

* also preserve audio tensors

* trailing space

* split arch A and V

* rm unused gemma3 func

* fix alignment

---------

Co-authored-by: Xuan Son Nguyen <son@huggingface.co>

tools/mtmd/clip-model.h

@@ -173,6 +173,45 @@ struct clip_layer {
     }
 };
 
+// Expanded MobileNetV5 block structure for Gemma3n vision encoder
+struct mobilenetv5_block {
+    // Stage 0 (Edge Residual)
+    ggml_tensor * s0_conv_exp_w = nullptr;
+    ggml_tensor * s0_bn1_w      = nullptr;
+    ggml_tensor * s0_conv_pwl_w = nullptr;
+    ggml_tensor * s0_bn2_w      = nullptr;
+
+    // Stage 1+ (Universal Inverted Residual)
+    ggml_tensor * dw_start_w    = nullptr;
+    ggml_tensor * dw_start_bn_w = nullptr;
+
+    ggml_tensor * pw_exp_w      = nullptr;
+    ggml_tensor * pw_exp_bn_w   = nullptr;
+
+    ggml_tensor * dw_mid_w      = nullptr;
+    ggml_tensor * dw_mid_bn_w   = nullptr;
+
+    ggml_tensor * pw_proj_w     = nullptr;
+    ggml_tensor * pw_proj_bn_w  = nullptr;
+
+    ggml_tensor * layer_scale_w = nullptr;
+
+    // Attention (MQA) components
+    ggml_tensor * attn_q_w = nullptr;
+    ggml_tensor * attn_k_w = nullptr;
+    ggml_tensor * attn_v_w = nullptr;
+    ggml_tensor * attn_o_w = nullptr;
+
+    // Optional downsampling/norm in attention
+    ggml_tensor * attn_k_dw_w   = nullptr;
+    ggml_tensor * attn_k_norm_w = nullptr;
+    ggml_tensor * attn_v_dw_w   = nullptr;
+    ggml_tensor * attn_v_norm_w = nullptr;
+
+    // Block norm (often present in attention blocks)
+    ggml_tensor * attn_norm_w   = nullptr;
+};
+
 struct clip_model {
     clip_modality modality = CLIP_MODALITY_VISION;
     projector_type proj_type = PROJECTOR_TYPE_MLP;
@@ -289,6 +328,23 @@ struct clip_model {
     ggml_tensor * mm_input_proj_w = nullptr;
     ggml_tensor * mm_soft_emb_norm_w = nullptr;
 
+    // mobilenetv5 for gemma3n
+    std::vector<mobilenetv5_block> mobilenet_blocks;
+    std::vector<int> mobilenet_stage_ends;
+    ggml_tensor * mobilenet_stem_conv_w = nullptr;
+    ggml_tensor * mobilenet_stem_conv_b = nullptr;
+    ggml_tensor * mobilenet_stem_norm_w = nullptr;
+    ggml_tensor * mm_post_proj_norm_w = nullptr;
+
+    // Multi-Scale Fusion Adapter (MSFA) components
+    ggml_tensor * msfa_concat_conv_w = nullptr;
+    ggml_tensor * msfa_concat_norm_w = nullptr;
+    ggml_tensor * msfa_ffn_expand_w = nullptr;
+    ggml_tensor * msfa_ffn_project_w = nullptr;
+    ggml_tensor * msfa_ffn_expand_bn = nullptr;
+    ggml_tensor * msfa_ffn_project_bn = nullptr;
+
+
     // pixtral, glm4v
     ggml_tensor * token_embd_img_break = nullptr;
     ggml_tensor * mm_patch_merger_w = nullptr;