{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreif4uwm77vyqkgcoeakuixqajaren43ffl4nvciyqkx2gzxfyf563y",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mgadi6dj2ps2"
  },
  "path": "/t/show-and-tell-m-courtyard-v0-4-9-local-mlx-fine-tuning-gui-for-mac-now-with-rslora-custom-jsonl/173995#post_1",
  "publishedAt": "2026-03-04T10:25:53.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "Kalajdzievski 2023",
    "https://github.com/Mcourtyard/m-courtyard"
  ],
  "textContent": "_**M-Courtyard v0.4.9** — Local LoRA fine-tuning on Apple Silicon (MLX)_\n\n_New in this release:_\n\n_**Custom JSONL Import (D-11)** : Import your own pre-formatted JSONL datasets directly from the UI. Supports **`{\"prompt\", \"completion\"}`** and **`{\"messages\": [...]}`** (chat) formats. Backend validates the first 5 lines for field completeness before accepting. If **`valid.jsonl`** is absent, the app auto-generates it from **`train.jsonl`** at training start._\n\n_**rsLoRA Scaling Strategy (T-3)** : New advanced option under the LoRA/DoRA training method. Switches from standard **`alpha/r`** to **`alpha/√r`** scaling — the approach introduced in **Kalajdzievski 2023**. Injects **`use_rslora: true`** into the **`lora_config.yaml`** passed to **`mlx_lm lora`**. Best for rank ≥ 32._\n\n_Built with: Tauri 2 + React + mlx-lm | macOS 14+ / Apple Silicon only_\n\n_GitHub:**https://github.com/Mcourtyard/m-courtyard**_\n\n",
  "title": "[Show and Tell] M-Courtyard v0.4.9: Local MLX Fine-Tuning GUI for Mac (Now with rsLoRA & Custom JSONL)"
}