Function clip_image_batch_encode

tools/mtmd/clip.cpp:3124–3782

}

bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_image_f32_batch * imgs_c_ptr, float * vec) {
    const clip_image_f32_batch & imgs = *imgs_c_ptr;
    int batch_size = imgs.entries.size();

    // TODO @ngxson : implement batch size > 1 as a loop
    //                we don't need true batching support because the cgraph will gonna be big anyway
    if (batch_size != 1) {
        return false; // only support batch size of 1
    }

    // if buffers are not allocated, we need to do a warmup run to allocate them
    if (!ctx->is_allocated) {
        clip_model_loader::warmup(*ctx, *imgs_c_ptr);
    }

    // build the inference graph
    ggml_backend_sched_reset(ctx->sched.get());
    ggml_cgraph * gf = clip_image_build_graph(ctx, imgs);
    ggml_backend_sched_alloc_graph(ctx->sched.get(), gf);

    // set inputs
    const auto & model   = ctx->model;
    const auto & hparams = model.hparams;

    const int image_size_width  = imgs.entries[0]->nx;
    const int image_size_height = imgs.entries[0]->ny;

    const int patch_size    = hparams.patch_size;
    const int num_patches   = ((image_size_width / patch_size) * (image_size_height / patch_size));
    const int n_pos = num_patches + (model.class_embedding ? 1 : 0);
    const int pos_w = image_size_width  / patch_size;
    const int pos_h = image_size_height / patch_size;


    auto get_inp_tensor = [&gf](const char * name) {
        ggml_tensor * inp = ggml_graph_get_tensor(gf, name);
        if (inp == nullptr) {
            GGML_ABORT("Failed to get tensor %s", name);
        }
        if (!(inp->flags & GGML_TENSOR_FLAG_INPUT)) {
            GGML_ABORT("Tensor %s is not an input tensor", name);
        }
        return inp;
    };

    auto set_input_f32 = [&get_inp_tensor](const char * name, std::vector<float> & values) {
        ggml_tensor * cur = get_inp_tensor(name);
        GGML_ASSERT(cur->type == GGML_TYPE_F32);
        GGML_ASSERT(ggml_nelements(cur) == (int64_t)values.size());
        ggml_backend_tensor_set(cur, values.data(), 0, ggml_nbytes(cur));
    };

    auto set_input_i32 = [&get_inp_tensor](const char * name, std::vector<int32_t> & values) {
        ggml_tensor * cur = get_inp_tensor(name);
        GGML_ASSERT(cur->type == GGML_TYPE_I32);
        GGML_ASSERT(ggml_nelements(cur) == (int64_t)values.size());
        ggml_backend_tensor_set(cur, values.data(), 0, ggml_nbytes(cur));
    };