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: 2
Learning rate: 5e-06
LoRA: r=128, alpha=128

Key Changes in 2c16

Based on 2c14 (best score 4.8386)
- All parameters maintained from 2c14
ALFWorld sample addition (+1)
- Clean & Place task pattern (2c14 had Heat/Cool but not Clean)
DBBench sample addition (+1)
- COUNT function improvement (was 27.3% in previous tests)
Upsampling adjustment
- ALF samples: 9 × 80x = 720
- DB samples: 6 × 55x = 330
- Total custom: 1,050

Usage

python

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

base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "TToyo2511/ttoyo_advance_2c16" #★TTT20260227 2c16版

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

Sources & Terms (IMPORTANT)

Training data:

u-10bei/sft_alfworld_trajectory_dataset_v5
u-10bei/dbbench_sft_dataset_react_v4 (5% sampled)

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.