https://github.com/ionden/mlx-quant-fidelity

Measure MLX quantization quality loss — KL divergence, perplexity, top-token agreement for KV cache and weights

apple-silicon diagnostics kl-divergence kv-cache kv-cache-quantization llm llm-eval llm-inference machine-learning metal mlx mlx-lm model-quantization perplexity python quantization quantization-quality

Last synced: 14 Jun 2026