Class ggml_context

tests/test_metal_win.cpp:27–27

    size_t ctx_size = ggml_tensor_overhead() * 16 + ggml_graph_overhead();
    struct ggml_init_params params = {ctx_size, nullptr, true};
    struct ggml_context* ctx = ggml_init(params);

    // Input tensor: [C, W, H, 1]
    auto* inp = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, C, W, H);