This repository provides a merged full model fine-tuned from
Qwen/Qwen3-4B-Instruct-2507 using LoRA + Unsloth.
The LoRA adapter has been merged into the base model weights.
The model can be loaded directly without requiring a separate adapter.
This model is trained for multi-turn agent-style reasoning tasks,
including structured tool use and database-oriented reasoning.
Loss is applied to all assistant turns within each trajectory.
The model was trained on a merged dataset created by concatenating
and shuffling the following datasets:
-
tussiiiii/agentbench_sft_mix_alfworld_dbbench_v1
A mixed dataset created from:
- u-10bei/sft_alfworld_trajectory_dataset_v5
- u-10bei/dbbench_sft_dataset_react_v4
-
tussiiiii/openalex_dbbench_synth_v3
-
tussiiiii/alfworld_synth_v1
-
tussiiiii/alfworld_synth_v2
The three datasets above were independently created by the author.
They are fully synthetic and were generated from scratch.
No benchmark evaluation data was used in their creation.
for making their datasets publicly available.
The synthetic datasets listed above were independently generated by the author.
This repository does not claim ownership of third-party datasets.
Synthetic datasets were independently generated.
1from transformers import AutoModelForCausalLM, AutoTokenizer
2import torch
3
4model_id = "your_id/your-repo"
5
6tokenizer = AutoTokenizer.from_pretrained(model_id)
7model = AutoModelForCausalLM.from_pretrained(
8 model_id,
9 torch_dtype=torch.bfloat16,
10 device_map="auto",
11)
