Modelpublic

hosyan/qwen3-4b-structured-output-slack-recovery-phase1-format-guard-t002

sourceHugging Faceapache-2.0updated 7mo agoView on Hugging Face

0likes12downloads

Model Card

qwen3-4b-structured-output-slack-recovery-phase1-format-guard-t002

This repository provides a LoRA adapter fine-tuned from Qwen/Qwen3-4B-Instruct-2507.

Training Configuration

—Method: QLoRA (4-bit, Unsloth)
—Max sequence length: 512
—Epochs: 3
—Learning rate: 2e-04
—LoRA: r=32, alpha=128, dropout=0.03

Usage

python

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

base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "hosyan/qwen3-4b-structured-output-slack-recovery-phase1-format-guard-t002"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(
    base,
    torch_dtype=torch.float16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)

Data

—Training dataset: u-10bei/structureddatawithcotdataset512v2
—Run name: slack-recovery-phase1-format-guard-t002