metadata
license: llama3
🔹 Key Highlights:
- 20% Fewer Parameters: nyun-c2-llama3-56B comprises approximately 20% fewer parameters than the popular Llama-3-70B.
- Better Performance: Despite having far fewer parameters, this model has better performance than Llama-3-70B.
- No Fine-Tuning Required: This model undergoes no fine-tuning, showcasing the raw potential of our optimization techniques.
Pipeline and Collaboration
For insights into the pipeline and the list of methods used to optimize these models, check out our PruneGPT repository (https://github.com/nyunAI/PruneGPT). We invite companies and organizations interested in joining forces with us to release more such open-source variants to reach out at [email protected].
Model Performance
| Dataset | nyun-c2-llama3-56B | Meta-Llama3-70B | Meta-Llama2-70B | MBZUAI K2-65B |
|---|---|---|---|---|
| MMLU (5-shot) | 78.4 | 79.5 | 69.7 | 67.9 |
| Winogrande (5-shot) | 85.5 | 83.1 | 81.8 | 77.0 |
| BoolQ (0-shot) | 85.1 | 79.0 | 73.1 | 83.0 |
| Hellaswag (10-shot) | 86.9 | 88.0 | 86.9 | 85.5 |
| Arc Challenge (25-shot) | 66.0 | 68.8 | 67.2 | 64.8 |
| GSM8K (5-shot) | 76.8 | 76.9 | 52.6 | 50.2 |
| Average | 79.8 | 79.2 | 71.9 | 71.4 |