Text Generation
MLX
English
next-concept-modeling
ncm
non-autoregressive
coarse-to-fine
apple-silicon
tinyworld
Eval Results (legacy)
Instructions to use janmusic/ncm-mlx-tinyworld with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use janmusic/ncm-mlx-tinyworld with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # if on a CUDA device, also pip install mlx[cuda] # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("janmusic/ncm-mlx-tinyworld") prompt = "Once upon a time in" text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use janmusic/ncm-mlx-tinyworld with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Generate some text mlx_lm.generate --model "janmusic/ncm-mlx-tinyworld" --prompt "Once upon a time"
- Atomic Chat
Download codec.json from janmusic/ncm-mlx-tinyworld: direct link, hf CLI and curl.
- Browser
- Download file 2.22 kB
-
https://huggingface.co/janmusic/ncm-mlx-tinyworld/resolve/main/codec.json
- Command line
-
hf download hf://janmusic/ncm-mlx-tinyworld/codec.json
-
curl -L -o codec.json https://huggingface.co/janmusic/ncm-mlx-tinyworld/resolve/main/codec.json
2.22 kB
| { | |
| "stage": "codec", | |
| "steps": 3000, | |
| "train_seconds": 52.21, | |
| "tokenizer": { | |
| "alphabet": " ._abcdefghiklnoprstuvwxy", | |
| "pad": "_", | |
| "seq_len": 32 | |
| }, | |
| "history": [ | |
| { | |
| "step": 250, | |
| "loss": 0.7816118597984314, | |
| "ce": 0.0183502696454525, | |
| "vq": 0.763261616230011 | |
| }, | |
| { | |
| "step": 500, | |
| "loss": 0.7766135931015015, | |
| "ce": 0.008087735623121262, | |
| "vq": 0.7685258388519287 | |
| }, | |
| { | |
| "step": 750, | |
| "loss": 1.9127604961395264, | |
| "ce": 1.155642032623291, | |
| "vq": 0.7571184635162354 | |
| }, | |
| { | |
| "step": 1000, | |
| "loss": 1.7280001640319824, | |
| "ce": 0.9738318920135498, | |
| "vq": 0.7541682720184326 | |
| }, | |
| { | |
| "step": 1250, | |
| "loss": 0.7600001096725464, | |
| "ce": 0.0029663536697626114, | |
| "vq": 0.7570337653160095 | |
| }, | |
| { | |
| "step": 1500, | |
| "loss": 1.7674469947814941, | |
| "ce": 1.0144059658050537, | |
| "vq": 0.7530410289764404 | |
| }, | |
| { | |
| "step": 1750, | |
| "loss": 1.321763038635254, | |
| "ce": 0.5726578235626221, | |
| "vq": 0.7491052746772766 | |
| }, | |
| { | |
| "step": 2000, | |
| "loss": 0.7995526790618896, | |
| "ce": 0.028942842036485672, | |
| "vq": 0.7706098556518555 | |
| }, | |
| { | |
| "step": 2250, | |
| "loss": 0.7662097811698914, | |
| "ce": 0.004673249088227749, | |
| "vq": 0.7615365386009216 | |
| }, | |
| { | |
| "step": 2500, | |
| "loss": 0.7780300974845886, | |
| "ce": 0.004726271145045757, | |
| "vq": 0.7733038067817688 | |
| }, | |
| { | |
| "step": 2750, | |
| "loss": 0.932603120803833, | |
| "ce": 0.16295668482780457, | |
| "vq": 0.7696464657783508 | |
| }, | |
| { | |
| "step": 3000, | |
| "loss": 0.7746670842170715, | |
| "ce": 0.0031447401270270348, | |
| "vq": 0.7715223431587219 | |
| } | |
| ], | |
| "val": { | |
| "char_acc_by_levels_used": [ | |
| 0.7497, | |
| 0.8812, | |
| 0.9568, | |
| 0.9865, | |
| 0.9942, | |
| 1.0 | |
| ], | |
| "seq_exact_by_levels_used": [ | |
| 0.0, | |
| 0.05, | |
| 0.366, | |
| 0.76, | |
| 0.878, | |
| 1.0 | |
| ], | |
| "codebook_utilization": [ | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0 | |
| ], | |
| "code_perplexity": [ | |
| 114.54, | |
| 106.31, | |
| 97.86, | |
| 104.75, | |
| 103.75, | |
| 96.55 | |
| ] | |
| } | |
| } | |