Note (reproducibility): Training was run with a 4-bit loading setup (QLoRA-style).
The adapter artifacts record unsloth/qwen3-4b-instruct-2507-unsloth-bnb-4bit as the base loading path in adapter_config.json, while the canonical base model ID for this run is Qwen/Qwen3-4B-Instruct-2507 as specified in meta.json. :contentReference[oaicite:1]{index=1} :contentReference[oaicite:2]{index=2}

This repository contains LoRA adapter weights only. The base model must be loaded separately.

Training Objective

This adapter is trained via supervised fine-tuning (SFT) to improve structured output accuracy (JSON / YAML / XML / TOML / CSV). :contentReference[oaicite:3]{index=3}

CoT masking: enabled (SFT_MASK_COT=1) :contentReference[oaicite:4]{index=4}
Output learn mode: after_marker (loss is applied only to assistant output after output markers) :contentReference[oaicite:5]{index=5}
Output markers:
- Output:, OUTPUT:, Final:, Answer:, Result:, Response: :contentReference[oaicite:6]{index=6}

Training Configuration

Key settings (from meta.json): :contentReference[oaicite:7]{index=7}

Base model: Qwen/Qwen3-4B-Instruct-2507
Dataset: u-10bei/structured_data_with_cot_dataset_512_v2
Max sequence length: 512
Epochs: 2
Learning rate: 2e-6
Weight decay: 0.01 (wd001)
LoRA: r=64, alpha=128, dropout=0
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Train batch size (per device): 2
Eval batch size (per device): 2
Gradient accumulation: 8
Eval steps: 50
Logging steps: 10
Save steps: 100
Save total limit: 2
Validation ratio: 0.05
Warmup ratio: 0.1
Seed: 3407
Precision: bf16 (USE_BF16=true)
Hardware (logged): NVIDIA A100-SXM4-80GB
Upsampling: disabled (SFT_USE_UPSAMPLING=0)
- (configured rule) {"xml_to_yaml": 2.0}

LoRA adapter config (from adapter_config.json): :contentReference[oaicite:8]{index=8}

Base loading path: unsloth/qwen3-4b-instruct-2507-unsloth-bnb-4bit
PEFT version: 0.18.1
LoRA: r=64, alpha=128, dropout=0.0
Target modules: down_proj, gate_proj, up_proj, k_proj, q_proj, v_proj, o_proj

Usage

python

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

# Canonical base model ID (from meta.json)
base = "Qwen/Qwen3-4B-Instruct-2507"

# Adapter repo ID
adapter = "your_id/qwen3-4b-sft-v0-cotv2-lr2e6-ep2-wd001"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(
    base,
    torch_dtype=torch.bfloat16,  # matches USE_BF16=true
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)

Sources & Terms (IMPORTANT)

Training data: u-10bei/structured_data_with_cot_dataset_512_v2

Dataset License/Terms: Please refer to the dataset card on Hugging Face and comply with its terms. Base Model Terms: You must comply with the base model’s original license/terms of use.

Repository License: MIT

sabia0080/qwen3-4b-sft-v0-cotv2-lr2e6-ep2-wd001

qwen3-4b-sft-v0-cotv2-lr2e6-ep2-wd001

Training Objective

Training Configuration

Usage

Sources & Terms (IMPORTANT)

sabia0080/
qwen3-4b-sft-v0-cotv2-lr2e6-ep2-wd001