LLM2025_comp_advanced_8

qwen2.5-7b-agent-trajectory-lora

This repository provides a LoRA adapter fine-tuned from Qwen/Qwen2.5-7B-Instruct using LoRA + Unsloth.

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

Training Objective

This adapter is trained to improve multi-turn agent task performance on ALFWorld (household tasks) and DBBench (database operations). Additional self-generated synthetic datasets are included to target known failure modes (e.g., SQL MAX aggregation and action efficiency).

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/Qwen2.5-7B-Instruct
• Method: LoRA (full precision base)
• Max sequence length: 2048
• Epochs: 8
• Learning rate: 1e-06
• LoRA: r=64, alpha=128

Usage

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

base = "Qwen/Qwen2.5-7B-Instruct"
adapter = "your_id/your-repo"

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 (organizer-provided): SFT_DATASET_DB , SFT_DATASET_ALF Additional training data (self-generated): synthetic_action_sft.cleaned.jsonl , synthetic_sql_maxfocus.jsonl , synthetic_sql_countfix.jsonl

Dataset License (organizer-provided datasets): MIT License. These organizer-provided datasets are used under the terms of the MIT License (including the copyright notice).

Synthetic data terms: The synthetic JSONL files are generated by the participant for this competition. Their use and redistribution (if any) follow the competition rules and applicable policies.

Compliance: Users must comply with the MIT License for organizer-provided datasets and the base model’s original terms of use.