Loss is applied to all assistant turns in the multi-turn trajectory, enabling the model to learn environment observation, action selection, tool use, and recovery from errors.

Training Configuration

Base model: Qwen/Qwen3-4B-Instruct-2507
Method: LoRA (full precision base)
Max sequence length: 2048
Epochs: 1
Learning rate: 2e-6
Warmup ratio: 0.05
LoRA: r=64, alpha=64
Target modules: q_proj, up_proj

Usage

With vLLM (Recommended for Docker Environment)

python

from vllm import LLM, SamplingParams

model = LLM(
    model="kevineen/qwen3-4b-alfworld-adv59-mix2048-neftune-lr2e6",
    max_model_len=8192,
    gpu_memory_utilization=0.95,
)

prompts = [
    "Your task is to: put a clean lettuce leaf in the fridge.",
    # ... more prompts
]

sampling_params = SamplingParams(temperature=0.0, max_tokens=512)
outputs = model.generate(prompts, sampling_params)

for output in outputs:
    print(output.outputs[0].text)

With Transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "kevineen/qwen3-4b-alfworld-adv59-mix2048-neftune-lr2e6"

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

# Inference
messages = [{"role": "user", "content": "Your task here"}]
inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt"
).to("cuda")
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0]))

Sources & Terms (IMPORTANT)

Training data: ALFWorld + DBBench mixed dataset

Datasets:

u-10bei/sft_alfworld_trajectory_dataset_v5
u-10bei/dbbench_sft_dataset_react_v4

Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License.

Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.