Skip to content

Fine-tuning documentation improvements#52

Merged
alay2shah merged 5 commits intomainfrom
alay2shah/finetuning-fixes
Feb 5, 2026
Merged

Fine-tuning documentation improvements#52
alay2shah merged 5 commits intomainfrom
alay2shah/finetuning-fixes

Conversation

@alay2shah
Copy link
Contributor

Summary

Comprehensive overhaul of the fine-tuning documentation with better structure, real dataset examples, and clearer guidance:

  • Add new Finetuning Datasets page covering SFT, DPO, VLM-SFT, and GRPO formats with example datasets
  • Add VLM-SFT section to TRL page with LoRA and full fine-tuning examples
  • Simplify Unsloth page to Getting Started guide linking to official docs
  • Replace dataset placeholders with real examples (smoltalk, orpo-dpo-mix, llava-instruct)
  • Update Colab links to point to GitHub main branch for long-term maintainability
  • Clean up notebooks: replace pip with uv pip, add consistent emoji headers

🤖 Generated with Claude Code

alay2shah and others added 5 commits February 5, 2026 09:59
- Update SFT, VLM-SFT, DPO notebooks to use uv pip
- Add emoji section headers to GRPO notebook for consistency
- Consolidate GRPO install commands

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add new datasets.mdx page covering SFT, DPO, VLM-SFT, GRPO formats
- Include example datasets for each training method
- Add dataset sources section (HuggingFace, JSONL, CSV, Parquet)
- Add PIL image loading example in accordion

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Vision Language Model Fine-Tuning (VLM-SFT) section with LoRA example
- Replace placeholder datasets with real examples (smoltalk, orpo-dpo-mix, llava-instruct)
- Update Colab links to point to GitHub main branch
- Add dataset format links to each training section

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Reduce from ~295 lines to ~90 lines
- Link to official Unsloth LFM2.5 documentation
- Add Unsloth Colab notebooks (SFT, GRPO, CPT)
- Keep essential Quick Start code example and tips
- Add dataset format link

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
@alay2shah alay2shah requested review from a team and Paulescu as code owners February 5, 2026 15:00
@alay2shah alay2shah merged commit e534168 into main Feb 5, 2026
6 checks passed
@alay2shah alay2shah deleted the alay2shah/finetuning-fixes branch February 5, 2026 22:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant