Datasetpublic

saki007ster/nutrition_dataset_mlx_lm

Dataset Structure The dataset is in JSONL format, with each line containing a JSON object with: text: The processed text chunk source: The source PDF file name Example format: {"text": "Nutrition is the science that interprets the interaction of nutrients and other substances in food...", "source": "Human-Nutrition-2020-Edition.pdf"} Using the Dataset with MLX-LM To use this dataset for fine-tuning the MLX-LM model: Make sure you have MLX-LM installed: pip… See the full description on the dataset page: https://huggingface.co/datasets/saki007ster/nutrition_dataset_mlx_lm.

sourceHugging Facemitupdated 1y agoView on Hugging Face

1likes20downloads

Dataset Card

Dataset Structure

The dataset is in JSONL format, with each line containing a JSON object with:

—text: The processed text chunk
—source: The source PDF file name

Example format:

json

{"text": "Nutrition is the science that interprets the interaction of nutrients and other substances in food...", "source": "Human-Nutrition-2020-Edition.pdf"}

Using the Dataset with MLX-LM

To use this dataset for fine-tuning the MLX-LM model:

1.Make sure you have MLX-LM installed:

bash

pip install mlx-lm

1.Use the dataset with MLX-LM's training script:

bash

python -m mlx_lm.train \
    --model mlx-community/Llama-3.2-1B-Instruct-4bit \
    --train \
    --data dataset/ \
    --batch-size 4 \
    --iters 1000 \
    --save-every 100 \
    --adapter-file adapter.npz