A community based open source LLM and what it can do for your business
summary
Claim: InstructLab is an opinionated fine‑tuning pipeline that uses small YAML Q&A stanzas to generate thousands of synthetic examples and teach a base model such as Granite 7B with as few as ~45 seed Qs. It runs locally (a full laptop training demo takes about five hours), the community build uses GitHub PRs and peer review and a server with eight A100s, supports document conversion via Dockling, reads tables and does basic math, and outputs versioned tuned models.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:15What is Instruct Lab
- 3:03Clarifying terms
- 4:09Fine-tuning vs RAG
- 6:12Instruct Lab workflow
- 8:20Demo: Mudang example
- 12:17Live model training
- 15:08Q&A session part one
- 17:14Q&A session part two
- 20:01Large documentation sets
- 22:07Automation vs human Q&A
- 23:33Documentation challenges
- 25:48Q&A session part three
- 26:10GitHub Actions CI/CD
- 27:53Closing remarks
related talks
40mInstructLab: Applying Open Source Methods to Building and Trai...- 40mInstructLab: Applying Open Source Methods to Building and Trai...
30mApplying Open Source Methods to Building and Training Large Langu...
59mWorkshop: Choosing the Best Open Source LLM for Your Application
01h 49mAI engineering with open access LLMs that lie, curse, and steal
05mKeynote: Teaching LLMs New Knowledge
50mPimp My Llama - How to Customize LLMs for Your Business Case
46mBeyond Basic Prompting: Supercharging Open Source LLMs with LMQL's Structured Generation