Quantization Benchmark
Run the same model in float32, float16, and int8 to compare inference speed, memory footprint, and prediction stability.
input image
No image selected yet.
Pick a preset or upload an image, then run the benchmark.
speed + memory
Run the benchmark to visualize speed and memory differences.
quality comparison
Quality metrics will appear after the benchmark runs.
weight inspector
Run the benchmark to inspect how values are rounded per precision.