Views
No views yet
| Parameter | Value |
|---|---|
| Base model | Qwen/Qwen2.5-7B-Instruct |
| Method | SDFT (On-Policy Self-Distillation) |
| Dataset | ToolAlpaca (4,046 training examples) |
| Training step | 700 / 1011 |
| Learning rate | 2e-5 (cosine schedule, 10% warmup) |
| Batch size | 32 (gradient accumulation) |
| Epochs | 1 |
| Precision | bf16 |
| Max prompt length | 1024 |
| Max completion length | 1024 |
| EMA alpha | 0.01 |
| Hardware | 1x NVIDIA L40S 48GB |
| Training time | ~42 hours (full run) |
| Metric | Base Model | This Model (Step 700) |
|---|---|---|
| Greedy Accuracy | 54.4% | 54.4% |
| pass@1 | 52.6% | 48.4% |
| pass@5 | 61.5% | 63.3% |
| pass@10 | 64.3% | 66.7% |
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model = AutoModelForCausalLM.from_pretrained("Ayushnangia/qwen2.5-7b-instruct-sdft-tooluse-step-700")
4tokenizer = AutoTokenizer.from_pretrained("Ayushnangia/qwen2.5-7b-instruct-sdft-tooluse-step-700")
5
6messages = [{"role": "user", "content": "Your tool-use prompt here"}]
7text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
8inputs = tokenizer(text, return_tensors="pt").to(model.device)
9outputs = model.generate(**inputs, max_new_tokens=1024)
10print(tokenizer.decode(outputs[0], skip_special_tokens=True))1@article{shenfeld2025selfdistillation,
2 title={Self-Distillation Enables Continual Learning},
3 author={Shenfeld, Idan and others},
4 journal={arXiv preprint arXiv:2601.19897},
5 year={2025}
6}