qwen3-4b-advanced-sft-v18-merged

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

• Dataset: deepkick/sft_alfworld_v5_action_format
• Method: LoRA SFT (merged into base model)
• Max sequence length: 4096
• Epochs: 1
• Learning rate: 1e-06
• LoRA: r=32, alpha=128

vLLM Compatibility

• LoRA adapter has been merged
• No tokenizer vocabulary modification
• Intended for AgentBench Advanced evaluation

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "deepkick/qwen3-4b-advanced-sft-v18-merged"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto"
)

Data / License Notes

• Dataset: deepkick/sft_alfworld_v5_action_format Please refer to the dataset page for license and terms.
• Base model: Qwen/Qwen3-4B-Instruct-2507 Please comply with the base model's original terms of use.