Qwen/Qwen2.5-0.5B-Instruct fine-tuned with mix of synthetic and real data.
- 1 epoch of SFT (5000 samples)
- Optimizer: PagedAdamW8bit
- Learning rate: 2e-5
- Batch size: 16
- Sample length: 1024 tokens
- Downloads last month
- 6
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support