minor
Browse files
README.md
CHANGED
|
@@ -1,3 +1,30 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
{}
|
| 3 |
+
---
|
| 4 |
+
# Introducing SmallThinker-3B: A Lightweight Model Fine-tuned on QwQ Synthetic Data
|
| 5 |
+
|
| 6 |
+
We introduce **SmallThinker-3B**, a new model fine-tuned from the [Qwen2.5-3b-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) model using synthetic data generated by [QwQ-32B-Preview](https://huggingface.co/Qwen/QwQ-32B-Preview).
|
| 7 |
+
|
| 8 |
+
## Benchmark Performance
|
| 9 |
+
|
| 10 |
+
| Model | AMPS_Hard Score |
|
| 11 |
+
|---------|----------------|
|
| 12 |
+
| SmallThinker | 58.0 |
|
| 13 |
+
| GPT-4o (2024-08-06) | 54.0 |
|
| 14 |
+
| Qwen2.5-3B-Instruct | 44.0 |
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
## Intended Use Cases
|
| 19 |
+
|
| 20 |
+
SmallThinker is designed for the following use cases:
|
| 21 |
+
|
| 22 |
+
1. **Edge Deployment:** Its small size makes it ideal for deployment on resource-constrained devices.
|
| 23 |
+
2. **Draft Model for QwQ-32B-Preview:** QwQ can serve as a fast and efficient draft model for the larger QwQ-32B-Preview model.
|
| 24 |
+
|
| 25 |
+
## Limitations & Disclaimer
|
| 26 |
+
|
| 27 |
+
Please be aware of the following limitations:
|
| 28 |
+
|
| 29 |
+
* **Language Limitation:** The model has only been trained on English-language datasets, hence its capabilities in other languages are still lacking.
|
| 30 |
+
* **Unpredictable Outputs:** The model may produce unexpected outputs due to its size and probabilistic generation paradigm. Users should exercise caution and validate the model's responses.
|