Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70B) with limited GPU memory, when you need to train <1% of parameters with minimal accuracy loss, or for multi-adapter serving. HuggingFace's official library integrated with transformers ecosystem.
77%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Fix and improve this skill with Tessl
tessl review fix ./cli-tool/components/skills/ai-research/fine-tuning-peft/SKILL.mdThe skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
The required workflow is a PEFT fine-tuning code example that calls `load_dataset` on a named Hugging Face Hub dataset (`databricks/databricks-dolly-15k`), a selected community-hosted item rather than a feed an outsider can push into, so the agent ingests outsider-authored text only after choosing a specific dataset.
The skill fetches instructions or code from an external URL at runtime, and the fetched content directly controls the agent’s prompts or executes code. This dynamic dependency allows the external source to modify the agent’s behavior without any changes to the skill itself.
The decisive boundary is the setup instruction in troubleshooting.md that clones and builds `TimDettmers/bitsandbytes` from GitHub and then runs `pip install .`; the repository is an individually owned account, so it is unofficial but reputable with no warning signs, which yields medium.
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.