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, which has strong performance in Math and Code with 15B activated params.
- 2/7/2024 - Serp-ai adds unsloth support for faster and memory efficient training of our Parameter-Efficient Sparsity Crafting and releases new sparsetral models based on mistral-7B.
- 1/10/2024 - Camelidae models are now available on 🤗 HuggingFace.
- 1/4/2024 - We release the paper, Parameter-Efficient Sparsity Crafting From Dense to Mixture-of-Experts for Instruction Tuning on General Tasks.
- 12/22/2023 - We release the training repo 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 and Adapter to perform Efficient Sparse Upcycling.
Model Lists
| Camelidae Series | Download |
|---|---|
| Camelidae-8x7B | 🤗 HuggingFace |
| Camelidae-8x13B | 🤗 HuggingFace |
| Camelidae-8x34B | 🤗 HuggingFace |
| Qwen2idae Series | Download |
|---|---|
| Qwen2idae-16x14B-v1.0 | 🤗 HuggingFace |
Performance
| Model | Activated Params | MMLU (5shot) | GSM8k (5shot) | MATH (4shot) | HumanEval (0shot) | MBPP (4shot) | HellaSwag (10shot) |
|---|---|---|---|---|---|---|---|
| GPT3.5 | - | 70.0% | 57.1% | 34.1% | 48.1% | - | 85.5% |
| LLaMA2-70B-chat | 70B | 63.8% | 59.3% | 10.4% | 32.3% | 35.6% | 84.8% |
| Camelidae-8x34B-pro | 35B | 75.7% | 79.4% | 24.0% | 48.8% | 43.2% | 85.2% |
| Camelidae-8x34B | 35B | 75.6% | 78.3% | 22.6% | 43.9% | 41.4% | 85.3% |
| SUSChat-34B | 34B | 76.4% | 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% | 77.8% | 29.9% | 62.8% | 48.6% | 82.3% |
| Mixtral-8x7B-instruct | 14B | 68.7% | 71.7% | 22.1% | 25.6% | 40.6% | 86.5% |
| 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
from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("hywu/Camelidae-8x7B", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("hywu/Camelidae-8x7B", device_map="auto", trust_remote_code=True).eval()
inputs = tokenizer('### Human:\nHow are you?\n### 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
@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. Camelidae models are developed for academic research and free commercial use, all usage must adhere to the license from facebookresearch and 01-ai.