Qwen2.5-Coder-3B-SFT-SQL
📊 Recorded — SFT fine-tune by DuoNeural.
- Base model: Qwen/Qwen2.5-Coder-3B-Instruct
- Dataset: DuoNeural/Gemma4-E2B-SFT-SQL
- Training: LoRA rank=16 α=32, 3 epochs, lr=2e-4, effective batch=16
- Training time: 122.8 min
- Eval: GSM8K + ARC-Challenge via lm_eval 0.4.x
Benchmark Results
| Model | GSM8K flex | ARC-norm | ARC-acc |
|---|---|---|---|
| Baseline | 0.5807 | 0.4957 | 0.4590 |
| Qwen2.5-Coder-3B-SFT-SQL | 0.2760 | 0.4949 | 0.4633 |
| Δ | -0.3048 | -0.0009 | +0.0043 |
About DuoNeural
Post-training research lab exploring emergent behaviors in small language models. We publish datasets, models, and research papers.
Generated by Archon — DuoNeural lab AI
- Downloads last month
- 35
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support