File size: 5,539 Bytes
0a24d74 e617865 0a24d74 e617865 0a24d74 e617865 0a24d74 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 |
---
datasets:
- Open-Orca/SlimOrca
- ise-uiuc/Magicoder-OSS-Instruct-75K
- ise-uiuc/Magicoder-Evol-Instruct-110K
- meta-math/MetaMathQA
language:
- en
library_name: transformers
pipeline_tag: text-generation
arxiv: 2401.02731
license: apache-2.0
---
# Parameter-Efficient Sparsity Crafting From Dense to Mixture-of-Experts for Instruction Tuning on General Tasks (EMNLP'24)
## News
- 9/20/2024 - Our paper is accepted by EMNLP'24.
- 3/12/2024 - We release Qwen2idae-16x14B-v1.0 on 🤗 [HuggingFace](https://huggingface.co/hywu/Qwen2idae-16x14B-v1.0), which has strong performance in Math and Code with 15B activated params.
- 2/7/2024 - [Serp-ai](https://github.com/serp-ai/Parameter-Efficient-MoE) adds [unsloth](https://github.com/serp-ai/unsloth) support for faster and memory efficient training of our Parameter-Efficient Sparsity Crafting and releases new [sparsetral](https://huggingface.co/serpdotai/sparsetral-16x7B-v2) models based on mistral-7B.
- 1/10/2024 - Camelidae models are now available on 🤗 [HuggingFace](https://huggingface.co/hywu).
- 1/4/2024 - We release the paper, [Parameter-Efficient Sparsity Crafting From Dense to Mixture-of-Experts for Instruction Tuning on General Tasks](https://arxiv.org/abs/2401.02731).
- 12/22/2023 - We release the training [repo](https://github.com/wuhy68/Parameter-Efficient-MoE) that craft the dense model with LLaMA architecture to the MoE model.
## Introduction
Camelidae and Qwen2idae models are trained utilizing Parameter-Efficient Sparsity Crafting techniques
We present Parameter-Efficient Sparsity Crafting to help dense models learn knowledge from different fields (including code and math). This approach performs instruction tuning and efficiently utilizes MoE structure.
Specifically, Parameter-Efficient Sparsity Crafting utilizes parameter-efficient techniques including [QLoRA](https://arxiv.org/abs/2305.14314) and [Adapter](https://arxiv.org/abs/1902.00751) to perform Efficient [Sparse Upcycling](https://arxiv.org/abs/2212.05055).
## Model Lists
| Camelidae Series | Download
|---|---
Camelidae-8x7B | 🤗 [HuggingFace](https://huggingface.co/hywu/Camelidae-8x7B)
Camelidae-8x13B | 🤗 [HuggingFace](https://huggingface.co/hywu/Camelidae-8x13B)
Camelidae-8x34B | 🤗 [HuggingFace](https://huggingface.co/hywu/Camelidae-8x34B)
Camelidae-8x34B-pro | 🤗 Coming Soon
| Qwen2idae Series | Download
|---|---
Qwen2idae-16x14B-v1.0 | 🤗 [HuggingFace](https://huggingface.co/hywu/Qwen2idae-16x14B-v1.0)
Qwen2idae-16x7B-v1.0 | 🤗 Coming Soon
Qwen2idae-16x1.8B-v1.0 | 🤗 Coming Soon
## Performance
| Model | Activated Params | MMLU (5shot) | GSM8k (5shot) | MATH (4shot) | HumanEval (0shot) | MBPP (4shot) | HellaSwag (10shot) |
|:-----:|:----------------:|:------------:|:-------------:|:------------:|:-----------------:|:------------:|:------------------:|
| GPT3.5 | - | 70.0% | 57.1% | <font color=#F67F70>**34.1%**</font> | <font color=#FBD98D>**48.1%**</font> | - | <font color=#7FEA9E>**85.5%**</font> |
| LLaMA2-70B-chat | 70B | 63.8% | 59.3% | 10.4% | 32.3% | 35.6% | 84.8% |
| Camelidae-8x34B-pro | 35B | <font color=#7FEA9E>**75.7%**</font> | <font color=#F67F70>**79.4%**</font> | <font color=#FBD98D>**24.0%**</font> | <font color=#7FEA9E>**48.8%**</font> | <font color=#7FEA9E>**43.2%**</font> | 85.2% |
| Camelidae-8x34B | 35B | <font color=#FBD98D>**75.6%**</font> | <font color=#7FEA9E>**78.3%**</font> | 22.6% | 43.9% | <font color=#FBD98D>**41.4%**</font> | <font color=#FBD98D>**85.3%**</font> |
| SUSChat-34B | 34B | <font color=#F67F70>**76.4%**</font> | 72.3% | 22.0% | 11.6% | 40.2% | 83.9% |
| Yi-34B-chat | 34B | 74.8% | 67.6% | 17.3% | 20.1% | 41.0% | 83.9% |
| Qwen2idae-16x14B-v1.0 | 15B | 66.7% | <font color=#FBD98D>**77.8%**</font> | <font color=#7FEA9E>**29.9%**</font> | <font color=#F67F70>**62.8%**</font> | <font color=#F67F70>**48.6%**</font> | 82.3% |
| Mixtral-8x7B-instruct | 14B | 68.7% | 71.7% | 22.1% | 25.6% | 40.6% | <font color=#F67F70>**86.5%**</font> |
| Camelidae-8x13B | 13B | 54.4% | 52.6% | 9.8% | 30.6% | 30.4% | 82.5% |
| LLaMA2-13B-chat | 13B | 53.9% | 37.1% | 5.2% | 18.9% | 27.2% | 81.9% |
| Camelidae-8x7B | 7B | 48.3% | 44.0% | 5.8% | 18.3% | 23.4% | 79.2% |
| LLaMA2-7B-chat | 7B | 47.2% | 26.3% | 3.9% | 12.2% | 17.6% | 78.6% |
We bold the top3 scores separately for all models.
## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("hywu/Qwen2idae-16x14B-v1.0", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("hywu/Qwen2idae-16x14B-v1.0", device_map="auto", trust_remote_code=True).eval()
inputs = tokenizer('<|im_start|>user\nHow are you?<|im_end|>\n<|im_start|>assistant\n', return_tensors='pt')
inputs = inputs.to(model.device)
pred = model.generate(**inputs)
print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
```
## Citation
```bibtex
@article{wu2024parameter,
title={Parameter-Efficient Sparsity Crafting from Dense to Mixture-of-Experts for Instruction Tuning on General Tasks},
author={Wu, Haoyuan and Zheng, Haisheng and Yu, Bei},
journal={arXiv preprint arXiv:2401.02731},
year={2024}
}
```
## License
The source code in this repo is licensed under the [Apache 2.0 License](https://github.com/wuhy68/Parameter-Efficient-MoE/blob/master/LICENSE). Qwen2idae models are developed for academic research and free commercial use, all usage must adhere to the license from [Qwen1.5](https://huggingface.co/Qwen/Qwen1.5-14B/blob/main/LICENSE). |