Menu
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/meridian-online/finetype --skill trainThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Showing SKILL.md
Based on SOC occupation classification
| name | train |
| description | Train a FineType CharCNN model with hardware auto-detection (Metal/CUDA/CPU) |
| user-invocable | true |
Run from the finetype repo root (~/github/noon-org/finetype/).
# Quick test (CPU, ~5 min)
./scripts/train.sh --samples 100 --size small --epochs 2
# Standard training
./scripts/train.sh --samples 1000 --size small --epochs 10
# Large model on M1 Mac (Metal auto-detected)
./scripts/train.sh --samples 5000 --size large --epochs 15 --seed 42
| Preset | embed_dim | num_filters | hidden_dim |
|---|---|---|---|
| small | 32 | 64 | 128 |
| medium | 64 | 128 | 256 |
| large | 128 | 256 | 512 |
Override individual params: --embed-dim 64 --num-filters 128 --hidden-dim 256
The script passes --features metal or --features cuda to Cargo automatically.
finetype generate --samples N)Output goes to models/char-cnn-vN/ (auto-incremented). Training log saved alongside.
--samples N Samples per type (default: 1000)
--size PRESET small|medium|large (default: small)
--epochs N Training epochs (default: 10)
--seed N Random seed (default: 42)
--embed-dim N Override embedding dimension
--num-filters N Override CNN filters
--hidden-dim N Override hidden layer dimension
--model-name NAME Output dir name (default: auto char-cnn-vN)
--data FILE Use existing NDJSON (skip generation)
./scripts/eval.sh --model models/char-cnn-vN # Evaluate
./scripts/package.sh models/char-cnn-vN # Package for distribution