Download c_engine/test_vision_encoder.cpp from aasivas/Parvulus: direct link, hf CLI and curl.
- Browser
- Download file 2.65 kB
-
https://huggingface.co/aasivas/Parvulus/resolve/main/c_engine/test_vision_encoder.cpp
- Command line
-
hf download hf://aasivas/Parvulus/c_engine/test_vision_encoder.cpp
-
curl -L -o test_vision_encoder.cpp https://huggingface.co/aasivas/Parvulus/resolve/main/c_engine/test_vision_encoder.cpp
2.65 kB
| int main() { | |
| std::cout << "=========================================================\n"; | |
| std::cout << " VISION ENCODER STANDALONE SELF-TEST & BENCHMARK\n"; | |
| std::cout << "=========================================================\n"; | |
| ParvulusEngine engine(0); | |
| if (!engine.load_weights("parvulus_model_weights.bin")) { | |
| std::cerr << "Failed to load parvulus_model_weights.bin\n"; | |
| return 1; | |
| } | |
| std::vector<float> h_obs(6 * 256 * 256, 0.5f); | |
| std::vector<float> h_text(256, 0.1f); | |
| float *d_obs, *d_text, *d_z_vis; | |
| cudaMalloc(&d_obs, 6 * 256 * 256 * sizeof(float)); | |
| cudaMalloc(&d_text, 256 * sizeof(float)); | |
| cudaMalloc(&d_z_vis, 16 * 256 * sizeof(float)); | |
| cudaMemcpy(d_obs, h_obs.data(), h_obs.size() * sizeof(float), cudaMemcpyHostToDevice); | |
| cudaMemcpy(d_text, h_text.data(), h_text.size() * sizeof(float), cudaMemcpyHostToDevice); | |
| std::cout << "[Test] Warming up Vision Backbone (20 iterations)...\n"; | |
| for (int i = 0; i < 20; ++i) { | |
| engine.forward_encoder(d_obs, d_text, d_z_vis); | |
| } | |
| cudaDeviceSynchronize(); | |
| std::cout << "[Test] Benchmarking 100 Vision Encoder iterations...\n"; | |
| auto t0 = std::chrono::high_resolution_clock::now(); | |
| for (int i = 0; i < 100; ++i) { | |
| engine.forward_encoder(d_obs, d_text, d_z_vis); | |
| } | |
| cudaDeviceSynchronize(); | |
| auto t1 = std::chrono::high_resolution_clock::now(); | |
| double avg_ms = std::chrono::duration<double, std::milli>(t1 - t0).count() / 100.0; | |
| std::cout << " -> Average Vision Encoder Latency: " << std::fixed << std::setprecision(3) << avg_ms << " ms\n"; | |
| std::vector<float> h_z_vis(16 * 256); | |
| cudaMemcpy(h_z_vis.data(), d_z_vis, 16 * 256 * sizeof(float), cudaMemcpyDeviceToHost); | |
| // Verify finite values and non-trivial output | |
| bool valid = true; | |
| for (float v : h_z_vis) { | |
| if (std::isnan(v) || std::isinf(v)) { | |
| valid = false; | |
| break; | |
| } | |
| } | |
| cudaFree(d_obs); | |
| cudaFree(d_text); | |
| cudaFree(d_z_vis); | |
| if (valid) { | |
| std::cout << " -> Vision output verified (finite, stable).\n"; | |
| std::cout << "=========================================================\n"; | |
| std::cout << " VISION ENCODER TEST PASSED!\n"; | |
| std::cout << "=========================================================\n"; | |
| return 0; | |
| } else { | |
| std::cerr << " -> Error: Output contains NaN or Inf!\n"; | |
| return 1; | |
| } | |
| } | |