Measure quantization quality loss on Apple Silicon MLX — KL divergence, top-token flip rate and perplexity delta for KV-cache and weight quantization
python machine-learning metal diagnostics quantization model-evaluation mlx kl-divergence perplexity model-quantization kv-cache apple-silicon llm local-llm llm-inference llm-eval kv-cache-quantization mlx-lm quantization-quality quantization-benchmark
-
Updated
Sep 6, 2026 - Python