|
--- |
|
datasets: |
|
- PowerInfer/QWQ-LONGCOT-500K |
|
- PowerInfer/LONGCOT-Refine-500K |
|
base_model: |
|
- Qwen/Qwen2.5-3B-Instruct |
|
pipeline_tag: text-generation |
|
language: |
|
- en |
|
library_name: transformers |
|
--- |
|
# SmallThinker-3B-preview |
|
|
|
We introduce **SmallThinker-3B-preview**, a new model fine-tuned from the [Qwen2.5-3b-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) model. |
|
|
|
## Benchmark Performance |
|
|
|
| Model | AIME24 | AMC23 | GAOKAO2024_I | GAOKAO2024_II | MMLU_STEM | AMPS_Hard | math_comp | |
|
|---------|--------|-------|--------------|---------------|-----------|-----------|-----------| |
|
| Qwen2.5-3B-Instruct | 6.67 | 45 | 50 | 35.8 | 59.8 | - | - | |
|
| SmallThinker | 16.667 | 57.5 | 64.2 | 57.1 | 68.2 | 70 | 46.8 | |
|
| GPT-4o | 9.3 | - | - | - | 64.2 | 57 | 50 | |
|
|
|
Limitation: Due to SmallThinker's current limitations in instruction following, for math_comp we adopt a more lenient evaluation method where only correct answers are required, without constraining responses to follow the specified AAAAA format. |
|
|
|
Colab Link: [Colab](https://colab.research.google.com/drive/182q600at0sVw7uX0SXFp6bQI7pyjWXQ2?usp=sharing) |
|
## Intended Use Cases |
|
|
|
SmallThinker is designed for the following use cases: |
|
|
|
1. **Edge Deployment:** Its small size makes it ideal for deployment on resource-constrained devices. |
|
2. **Draft Model for QwQ-32B-Preview:** SmallThinker can serve as a fast and efficient draft model for the larger QwQ-32B-Preview model. From my test, in llama.cpp we can get 70% speedup (from 40 tokens/s to 70 tokens/s). |
|
|
|
## Training Details |
|
|
|
The model was trained using 8 H100 GPUs with a global batch size of 16. The specific configuration is as follows: |
|
|
|
The SFT (Supervised Fine-Tuning) process was conducted in two phases: |
|
|
|
1. First Phase: |
|
- Used only the PowerInfer/QWQ-LONGCOT-500K dataset |
|
- Trained for 1.5 epochs |
|
``` |
|
### model |
|
model_name_or_path: /home/syx/Qwen2.5-3B-Instruct |
|
|
|
### method |
|
stage: sft |
|
do_train: true |
|
finetuning_type: full |
|
deepspeed: examples/deepspeed/ds_z3_config.json |
|
|
|
### dataset |
|
dataset: o1-v2 |
|
template: qwen |
|
neat_packing: true |
|
cutoff_len: 16384 |
|
overwrite_cache: true |
|
preprocessing_num_workers: 16 |
|
|
|
### output |
|
output_dir: saves/qwen2-01-qat/full/sft |
|
logging_steps: 1 |
|
save_steps: 1000 |
|
plot_loss: true |
|
overwrite_output_dir: true |
|
``` |
|
2. Second Phase: |
|
- Combined training with PowerInfer/QWQ-LONGCOT-500K and PowerInfer/LONGCOT-Refine datasets |
|
- Continued training for 2 additional epochs |
|
``` |
|
### model |
|
model_name_or_path: saves/qwen2-01-qat/full/sft/checkpoint-24000 |
|
|
|
### method |
|
stage: sft |
|
do_train: true |
|
finetuning_type: full |
|
deepspeed: examples/deepspeed/ds_z3_config.json |
|
|
|
### dataset |
|
dataset: o1-v2, o1-v3 |
|
template: qwen |
|
neat_packing: true |
|
cutoff_len: 16384 |
|
overwrite_cache: true |
|
preprocessing_num_workers: 16 |
|
|
|
### output |
|
output_dir: saves/qwen2-01-qat/full/sft |
|
logging_steps: 1 |
|
save_steps: 1000 |
|
plot_loss: true |
|
overwrite_output_dir: true |
|
``` |
|
|
|
## Limitations & Disclaimer |
|
|
|
Please be aware of the following limitations: |
|
|
|
* **Language Limitation:** The model has only been trained on English-language datasets, hence its capabilities in other languages are still lacking. |
|
* **Limited Knowledge:** Due to limited SFT data and the model's relatively small scale, its reasoning capabilities are constrained by its knowledge base. |
|
* **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. |
|
* **Repetition Issue:** The model tends to repeat itself when answering high-difficulty questions. Please increase the `repetition_penalty` to mitigate this issue. |
