Infinity-Parser2-Pro Int8

This repository contains a quantized derivative of infly/Infinity-Parser2-Pro.
The base model is distributed by INF Tech under the Apache License, Version 2.0.

Changes:
- Converted eligible language-model linear weights to MLX affine int8.
- Used group size 64 for quantized weights.
- Adapted fused expert tensor naming/layout for the native mere.run Q35 runtime.
- Added mere.run model metadata.