JerrySiRi/Qwen3-8B-lora-tulu-sft

qwen3-8B-32-64-5k

PEFT LoRA adapter fine-tuned from Qwen/Qwen3-8B on rl-research/dr-tulu-sft-data.

Training Details

• LoRA rank: 32, alpha: 64
• Target modules: q_proj, v_proj, k_proj, up_proj, down_proj, gate_proj, o_proj
• Trained with LlamaFactory on 2x GPUs, 3 epochs, cosine LR schedule.

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model_id = "Qwen/Qwen3-8B"
base = AutoModelForCausalLM.from_pretrained(base_model_id, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
model = PeftModel.from_pretrained(base, "qwen3-8B-32-64-5k")