# PsiLM on Qwen3.5 9B -- Apple Silicon (MLX). Verified 2026-09-11 on macOS / Python 3.11 # with: mlx 0.32.2, mlx-lm 0.31.3, transformers 5.16.1, torch 2.13.0, huggingface_hub 1.29.0. # # pip install -r requirements.txt # python psilm_infer.py # mlx>=0.32.2 # the first MLX that reads NVFP4 natively mlx-lm==0.31.3 # exact: the Qwen3.5 loader (psilm.mlx.qwen35_loader) wraps # mlx_lm.models.qwen3_5, which this version ships; others untested transformers>=4.51 # chat template only (apply_chat_template); the model runs in MLX torch>=2.4 # imported by psilm.stage2.qa (batching helpers); CPU wheel is fine huggingface_hub>=0.30 # downloads the backbone from ryoji-info/Qwen3.5-9B-nvfp4-mlx numpy>=1.26 safetensors>=0.4 # The PsiLM package itself (bridges, staged forward, the Qwen3.5 loader, FNO, QA builder). # Alternative to this line: clone the repository and set PSILM_REPO=/path/to/PsiLM. psilm @ git+https://github.com/ryoji-info/PsiLM