spb/zyquo-mlx Public MIT
The local MLX foundry for your Mac — run, fine-tune, quantize, and ship models. Nothing leaves your machine.
Swift 93.4%
Python 3.8%
Makefile 2.2%
Shell 0.5%
-
phase3: checkpoint — gate green (QLoRA loss 4.62→0.10, fuse w/ auto-dequantize, style transfer verified, Swift quant 0.1%-accurate preview); callback + requantization landmines documented
…
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
-
phase3: foundry core — PyBridge (PythonRunner, uv venv env, JSON-protocol scripts), DatasetService, TrainingService/RunStore/MetricsStream, ConversionService (Swift quant + Py fuse/convert), CLI drivers
…
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>