Hugging Peers

graphite-001-large

☆ Save
by appvoid text-generation transformers
19 downloads 0 likes on Hugging Face updated

Model card

graphite-001

Private fine-tune of ibm-granite/granite-4.0-350m on appvoid/rewrite.

Training

  • Method: full supervised fine-tuning
  • Epochs: 1
  • Dataset fields: instruction, text, output
  • Format: native chat template from the base tokenizer
  • Train rows used: 33609

Prompt format

The model was trained with one user message:

{instruction}

Text:
{text}

Return only the final rewritten output. Do not explain.

The assistant message contains only:

{output}

Intended use

Text rewriting, paraphrasing, tone transfer, grammar-style editing, and instruction-following text transformations.