Function sam3_load_hparams

sam3.cpp:1481–1524

*****************************************************************************/

static bool sam3_load_hparams(std::ifstream& fin, sam3_hparams& hp) {
    auto rd = [&](int32_t& v) { fin.read(reinterpret_cast<char*>(&v), 4); };
    rd(hp.img_size);
    rd(hp.patch_size);
    rd(hp.vit_embed_dim);
    rd(hp.vit_depth);
    rd(hp.vit_num_heads);
    int32_t mlp_ratio_x1000;
    rd(mlp_ratio_x1000);
    hp.vit_mlp_dim = static_cast<int32_t>(hp.vit_embed_dim * (mlp_ratio_x1000 / 1000.0f));
    rd(hp.vit_window_size);
    rd(hp.n_global_attn);
    for (int i = 0; i < hp.n_global_attn && i < 4; ++i) {
        rd(hp.global_attn_idx[i]);
    }
    rd(hp.text_width);
    rd(hp.text_heads);
    rd(hp.text_layers);
    rd(hp.text_ctx_len);
    rd(hp.text_vocab_size);
    rd(hp.text_out_dim);
    rd(hp.neck_dim);
    rd(hp.fenc_layers);
    rd(hp.fenc_heads);
    rd(hp.fenc_ffn_dim);
    rd(hp.ddec_layers);
    rd(hp.ddec_heads);
    rd(hp.ddec_ffn_dim);
    rd(hp.ddec_num_queries);
    rd(hp.geom_layers);
    rd(hp.n_presence_tokens);
    rd(hp.n_geom_queries);
    rd(hp.sam_embed_dim);
    rd(hp.sam_dec_depth);
    rd(hp.sam_n_multimask);
    rd(hp.sam_iou_head_depth);
    rd(hp.mem_out_dim);
    rd(hp.mem_attn_layers);
    rd(hp.num_maskmem);
    rd(hp.max_obj_ptrs);
    rd(hp.n_amb_experts);
    rd(hp.visual_only);
    return !fin.fail();
}

static void sam3_print_hparams(const sam3_hparams& hp) {
    fprintf(stderr, "  img_size       = %d\n", hp.img_size);