model : support step3-vl-10b (#21287)

* feat: support step3-vl-10b * use fused QKV && mapping tensor in tensor_mapping.py * guard hardcoded params and drop crop metadata * get understand_projector_stride from global config * img_u8_resize_bilinear_to_f32 move in step3vl class * Apply suggestions from code review Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * fix the \r\n mess * add width and heads to MmprojModel.set_gguf_parameters --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
2026-04-08 15:51:31 +08:00
parent 97508acb17
commit 09343c0198
12 changed files with 537 additions and 4 deletions
@@ -0,0 +1,81 @@
+#include "models.h"
+
+ggml_cgraph * clip_graph_step3vl::build() {
+    GGML_ASSERT(model.class_embedding == nullptr);
+    GGML_ASSERT(model.patch_embeddings_0 != nullptr);
+    GGML_ASSERT(model.position_embeddings != nullptr);
+
+    norm_type norm_t = NORM_TYPE_NORMAL;
+
+    ggml_tensor * pos_h = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches);
+    ggml_set_name(pos_h, "pos_h");
+    ggml_set_input(pos_h);
+
+    ggml_tensor * pos_w = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches);
+    ggml_set_name(pos_w, "pos_w");
+    ggml_set_input(pos_w);
+
+    ggml_tensor * inp = build_inp();
+    ggml_tensor * learned_pos_embd = resize_position_embeddings();
+
+    auto add_pos = [&](ggml_tensor * cur, const clip_layer &) {
+        return build_rope_2d(ctx0, cur, pos_w, pos_h, hparams.rope_theta, false);
+    };
+
+    auto add_spatial_bias = [&](ggml_tensor * cur, ggml_tensor * bias) {
+        if (bias == nullptr) {
+            return cur;
+        }
+
+        const int64_t width    = cur->ne[0];
+        const int64_t height   = cur->ne[1];
+        const int64_t channels = cur->ne[2];
+
+        cur = ggml_reshape_2d(ctx0, cur, width * height, channels);
+        cur = ggml_cont(ctx0, ggml_transpose(ctx0, cur));
+        cur = ggml_add(ctx0, cur, bias);
+        cur = ggml_cont(ctx0, ggml_transpose(ctx0, cur));
+        cur = ggml_reshape_3d(ctx0, cur, width, height, channels);
+
+        return cur;
+    };
+
+    ggml_tensor * cur = build_vit(
+        inp,
+        n_patches,
+        norm_t,
+        hparams.ffn_op,
+        learned_pos_embd,
+        add_pos);
+    cb(cur, "vit_out", -1);
+
+    // [n_embd, n_patches] -> [w, h, n_embd] for spatial downsampling convolutions.
+    cur = ggml_permute(ctx0, cur, 1, 0, 2, 3);
+    cur = ggml_cont_3d(ctx0, cur, n_patches_x, n_patches_y, n_embd);
+
+    // First downsampler: Conv2d(1536 -> 3072, k=3, s=2, p=1)
+    cur = ggml_conv_2d(ctx0, model.mm_0_w, cur, 2, 2, 1, 1, 1, 1);
+    cur = add_spatial_bias(cur, model.mm_0_b);
+    cb(cur, "downsample_0", -1);
+
+    // Second downsampler: Conv2d(3072 -> 6144, k=3, s=2, p=1)
+    cur = ggml_conv_2d(ctx0, model.mm_1_w, cur, 2, 2, 1, 1, 1, 1);
+    cur = add_spatial_bias(cur, model.mm_1_b);
+    cb(cur, "downsample_1", -1);
+
+    // [w, h, c] -> [c, w*h]
+    {
+        const int64_t w = cur->ne[0];
+        const int64_t h = cur->ne[1];
+        cur = ggml_reshape_3d(ctx0, cur, w * h, cur->ne[2], cur->ne[3]);
+        cur = ggml_cont(ctx0, ggml_permute(ctx0, cur, 1, 0, 2, 3));
+    }
+    cb(cur, "downsample_flatten", -1);
+
+    // Final projector: Linear(6144 -> projection_dim)
+    cur = ggml_mul_mat(ctx0, model.mm_model_proj, cur);
+    cb(cur, "projector_out", -1);
+
+    ggml_build_forward_expand(gf, cur);
+    return gf;
+}