peft-fine-tuning
Fine-tune large language models with less memory.
What it does
Guides fine-tuning large language models into small, task-specific adapters while limiting the number of trained parameters.
When to use it
Use for fine-tuning large models with limited GPU memory, training a small fraction of parameters, or serving multiple task-specific adapters.
How to use it
It changes how the agent implements model fine-tuning, configuring adapters, conserving memory, and validating results before merging.
What you provide
- An existing project
Uses
What you need · 2
The guide is about the peft library and declares peft>=0.13.0.
The documented large-model training workflows use GPUs, with RTX 4090 and A100 given as examples.
Access · 2
This skill
meta-llama/Llama-3.1-8B
Downloads the base language model
databricks/databricks-dolly-15k
Downloads the training dataset