Fine-Tuning LLMs

LLM

Fine-Tuning

Hugging Face

A short, practical walkthrough of fine-tuning a large language model with LoRA instead of full fine-tuning.

Published

January 20, 2026

Fine-Tuning LLMs

Pretrained LLMs are general-purpose. Fine-tuning adapts one to a narrower domain or task using far less data and compute than pretraining from scratch.

Full Fine-Tuning vs PEFT

Full fine-tuning updates every weight in the model — accurate, but expensive in GPU memory and easy to overfit on small datasets. Parameter-efficient fine-tuning (PEFT) methods like LoRA freeze the base model and train small low-rank adapter matrices instead, cutting trainable parameters by orders of magnitude while keeping most of the quality.

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import LoraConfig, get_peft_model

model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3-8b")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3-8b")

lora_config = LoraConfig(
    r=16,
    lora_alpha=32,
    target_modules=["q_proj", "v_proj"],
    lora_dropout=0.05,
    task_type="CAUSAL_LM",
)

model = get_peft_model(model, lora_config)
model.print_trainable_parameters()

Data Preparation

Keep examples in a consistent instruction/response format.
Favor quality over quantity — a few thousand well-curated examples usually beat a large noisy dataset.
Deduplicate aggressively; repeated examples skew the model more than expected.

Evaluation

Track validation loss, but don’t stop there — run side-by-side qualitative comparisons against the base model on real prompts. Loss going down doesn’t always mean the outputs are actually better for your task.

Fine-tuning is iterative: start small with LoRA on a subset of data before committing to a full fine-tune or a larger training run.