book-sft-pipeline
This skill should be used when the user asks to "fine-tune on books", "create SFT dataset", "train style model", "extract ePub text", or mentions style transfer, LoRA training, book segmentation, or author voice replication.
Security Assessment
About book-sft-pipeline
The book-sft-pipeline skill is designed to facilitate the process of fine-tuning machine learning models on literary works. It effectively converts raw ePub files into structured datasets that can be used for training style-transfer models. By addressing the challenges associated with extracting and segmenting text from books, this skill significantly enhances the quality of training data, thereby enabling models to replicate an author's unique voice and style. This skill is particularly valuable for those looking to create models that can produce text in the style of specific authors or literary genres.
FAQ
How do I activate the book-sft-pipeline skill?
Activate this skill when you need to fine-tune datasets from literary works or create author-voice models.
What formats does the skill support for text extraction?
The skill primarily supports ePub format, as it avoids the OCR errors associated with PDF files.
What are the limitations of the segmentation phase?
The segmentation phase works best with text chunks of 150-400 words to ensure semantic coherence and effective training.
Install book-sft-pipeline
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill