How do I install the Train Mlx skill?

Run npx skillmds@latest add mybigday/train-mlx in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Train Mlx safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Train Mlx?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Train Mlx free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Train Mlx?

mybigday (@mybigday) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Train Mlx by mybigday — Agent Skill

Name: Train Mlx
Author: mybigday

Train Mlx

LoRA fine-tune a causal LM natively on Apple Silicon (M1/M2/M3/M4) via `mlx-lm`. Triggered when the user is on macOS arm64, mentions MLX, or wants better throughput than torch-MPS. Use this instead of `train-sft` when the host is a Mac with no NVIDIA GPU.

mybigday 41adca2 2.5 KB Updated Sep 17, 2026

File contents

train-mlx — Apple MLX LoRA fine-tune

For everything except Apple Silicon, prefer the train-sft skill (torch + TRL). MLX wins on M-series because the unified memory means a 7B model that spills VRAM on a 24GB consumer NVIDIA card fits comfortably on a 36GB M-series.

When to fire

uname -sm returns Darwin arm64.
User mentions MLX, M-series, "Mac GPU", or unified memory.
Torch-MPS keeps OOM-ing or running out of MPS-supported ops.

Sequence

Install the MLX path (idempotent):

make env-mlx        # bootstrap with --accel=mps --mlx

Get the data into MLX's JSONL layout:
```
python scripts/dump_mlx_jsonl.py trl-lib/Capybara data/capybara-mlx
```
Produces train.jsonl + valid.jsonl with the messages (or text) field. You can skip this if you already have HF JSONL on disk.
Pick a 4-bit MLX-community base for fast loading on 16-32GB Macs:
- mlx-community/SmolLM2-1.7B-Instruct-4bit
- mlx-community/Llama-3.2-3B-Instruct-4bit
- mlx-community/Qwen2.5-7B-Instruct-4bit

Smoke run, then scale:

python scripts/mlx_finetune.py --config configs/mlx_default.yaml --iters 50
python scripts/mlx_finetune.py --config configs/mlx_default.yaml --iters 1000 --fuse

--fuse merges the LoRA back into the base weights so you can load the result with plain mlx_lm.generate or convert back to safetensors.

Push to Hub (optional): mlx_lm.fuse --upload-repo you/your-mlx-model.

Memory budget rules of thumb

Model	Min unified memory	Notes
≤1.7B	16 GB	works on M1/M2 base
3B	24 GB	M2 Pro / M3 Pro
7B 4-bit	36 GB	M3 Max / M4 Pro
13B 4-bit	64 GB	M2 Ultra / M3 Max

What MLX does NOT support

DPO/GRPO out of the box (mlx-lm has SFT + LoRA + DoRA — for preference data, generate offline preferences and do SFT-on-chosen, or use torch-MPS via train-dpo skill at slower speeds).
Multi-GPU (single-Mac only — fine, since you have one accelerator anyway).
bitsandbytes-style 8-bit optimizers (MLX has its own quantization).

Eval

scripts/eval_lm.py (lm-eval-harness) loads MLX-fused safetensors fine. For in-process MLX eval: mlx_lm.generate --model outputs/mlx-lora-fused --prompt "...".

---name: train-mlxdescription: LoRA fine-tune a causal LM natively on Apple Silicon (M1/M2/M3/M4) via `mlx-lm`. Triggered when the user is on macOS arm64, mentions MLX, or wants better throughput than torch-MPS. Use this instead of `train-sft` when the host is a Mac with no NVIDIA GPU.---# train-mlx — Apple MLX LoRA fine-tuneFor everything except Apple Silicon, prefer the `train-sft` skill (torch +TRL). MLX wins on M-series because the unified memory means a 7B model thatspills VRAM on a 24GB consumer NVIDIA card fits comfortably on a 36GB M-series.## When to fire- `uname -sm` returns `Darwin arm64`.- User mentions MLX, M-series, "Mac GPU", or unified memory.- Torch-MPS keeps OOM-ing or running out of MPS-supported ops.## Sequence1. **Install the MLX path** (idempotent): ```bash make env-mlx # bootstrap with --accel=mps --mlx ```2. **Get the data into MLX's JSONL layout**: ```bash python scripts/dump_mlx_jsonl.py trl-lib/Capybara data/capybara-mlx ``` Produces `train.jsonl` + `valid.jsonl` with the `messages` (or `text`) field. You can skip this if you already have HF JSONL on disk.3. **Pick a 4-bit MLX-community base** for fast loading on 16-32GB Macs: - `mlx-community/SmolLM2-1.7B-Instruct-4bit` - `mlx-community/Llama-3.2-3B-Instruct-4bit` - `mlx-community/Qwen2.5-7B-Instruct-4bit`4. **Smoke run, then scale**: ```bash python scripts/mlx_finetune.py --config configs/mlx_default.yaml --iters 50 python scripts/mlx_finetune.py --config configs/mlx_default.yaml --iters 1000 --fuse ``` `--fuse` merges the LoRA back into the base weights so you can load the result with plain `mlx_lm.generate` or convert back to safetensors.5. **Push to Hub** (optional): `mlx_lm.fuse --upload-repo you/your-mlx-model`.## Memory budget rules of thumb| Model | Min unified memory | Notes ||---|---|---|| ≤1.7B | 16 GB | works on M1/M2 base || 3B | 24 GB | M2 Pro / M3 Pro || 7B 4-bit | 36 GB | M3 Max / M4 Pro || 13B 4-bit | 64 GB | M2 Ultra / M3 Max |## What MLX does NOT support- DPO/GRPO out of the box (mlx-lm has SFT + LoRA + DoRA — for preference data, generate offline preferences and do SFT-on-chosen, or use torch-MPS via `train-dpo` skill at slower speeds).- Multi-GPU (single-Mac only — fine, since you have one accelerator anyway).- `bitsandbytes`-style 8-bit optimizers (MLX has its own quantization).## Eval`scripts/eval_lm.py` (lm-eval-harness) loads MLX-fused safetensors fine. Forin-process MLX eval: `mlx_lm.generate --model outputs/mlx-lora-fused--prompt "..."`.

Frequently asked questions