Function test_sinusoidal_pe

tests/test_vit.cpp:234–290

}

static bool test_sinusoidal_pe(const std::string & ref_dir) {
    fprintf(stderr, "\n=== Test: Sinusoidal PE ===\n");

    struct { int H; int W; const char * name; } cases[] = {
        {288, 288, "pe_288"},
        {144, 144, "pe_144"},
        { 72,  72, "pe_72"},
        { 36,  36, "pe_36"},
    };

    for (auto & tc : cases) {
        auto ref = load_ref(ref_dir + "/" + tc.name);
        if (ref.data.empty()) continue;

        // Compute our sinusoidal PE
        const int H = tc.H, W = tc.W, d_model = 256;
        const int half = d_model / 2;
        const float scale = 2.0f * (float)M_PI;
        const float temperature = 10000.0f;

        // ref is [1, 256, H, W] in NCHW from PyTorch
        std::vector<float> our_pe(d_model * H * W);

        for (int y = 0; y < H; ++y) {
            for (int x = 0; x < W; ++x) {
                float pos_y = ((float)(y + 1) / (float)(H)) * scale;
                float pos_x = ((float)(x + 1) / (float)(W)) * scale;

                for (int i = 0; i < half; ++i) {
                    float dim_t = powf(temperature, 2.0f * (float)(i / 2) / (float)half);

                    float val_x, val_y;
                    if (i % 2 == 0) {
                        val_x = sinf(pos_x / dim_t);
                        val_y = sinf(pos_y / dim_t);
                    } else {
                        val_x = cosf(pos_x / dim_t);
                        val_y = cosf(pos_y / dim_t);
                    }

                    // PyTorch output is [1, 256, H, W] = [N, C, H, W]
                    // Channel layout: first half is pos_y, second half is pos_x
                    // Index: [0, c, y, x] → c * H * W + y * W + x
                    our_pe[i * H * W + y * W + x] = val_y;
                    our_pe[(i + half) * H * W + y * W + x] = val_x;
                }
            }
        }

        float max_diff = compare_tensors(our_pe.data(), ref.data.data(), d_model * H * W);
        bool ok = max_diff < 1e-5f;
        fprintf(stderr, "  %s %s (%dx%d): max_diff=%.6e\n",
                ok ? "[PASS]" : "[FAIL]", tc.name, tc.H, tc.W, max_diff);
    }

    return true;
}