llama32-1b-qa-lora

LoRA adapter (PEFT) — Llama 3.2 1B QA tutor

Repository: NeuralVulture/llama32-1b-qa-lora
Base model: unsloth/Llama-3.2-1B-Instruct
Uploaded: 2026-03-26 (UTC)

Contents

This repo contains a LoRA adapter only (low-rank weights + adapter_config.json).
You must load it on top of the base model with PEFT / Transformers.

Merged full weights (single from_pretrained, no PEFT):
https://huggingface.co/NeuralVulture/llama32-1b-qa-merged

Load (PEFT)

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_id = "unsloth/Llama-3.2-1B-Instruct"
adapter_id = "NeuralVulture/llama32-1b-qa-lora"

tokenizer = AutoTokenizer.from_pretrained(adapter_id)
base = AutoModelForCausalLM.from_pretrained(
    base_id,
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True,
)
model = PeftModel.from_pretrained(base, adapter_id)

Training

Adapter exported from local LoRA SFT (TRL + PEFT). Local source directory (reference): /workspace/outputs/lora-llama-3.2-1b-qa.

License

The base model license applies to the base weights. This adapter is a derivative; use only in compliance with the base model license and your jurisdiction.