Downloads · 30 days
38
1% of all-time downloads
JunxiongWang/Llama3.2-Mamba2-3B-distill
Llama3.2-Mamba2-3B-distill is a machine learning model from JunxiongWang. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Zero-shot results when using the Llama-3.1-70B-Instruct as the teacher model, and the Llama-3.2-3B-Instruct as the initialized model
Downloads · 30 days
38
1% of all-time downloads
All-time downloads
7.2K
Public
Repo size
26.8 GB
Likes
0
Public
Click a slice to open those files.
.bin13.4 GB · 100%
From the Hugging Face model README
Zero-shot results when using the Llama-3.1-70B-Instruct as the teacher model, and the Llama-3.2-3B-Instruct as the initialized model
| Model | Llama-3.2-3B-Instruct | Llama3.2-Mamba-3B-distill | Llama3.2-Mamba-3B-dpo | Llama3.2-Mamba2-3B-distill | Llama3.2-Mamba2-3B-dpo |
|---|---|---|---|---|---|
| Initialization Model | N/A | Llama-3.2-3B-Instruct | Llama-3.2-3B-Instruct | Llama-3.2-3B-Instruct | Llama-3.2-3B-Instruct |
| Teacher Model | N/A | Llama-3.1-70B-Instruct | Llama-3.1-70B-Instruct | Llama-3.1-70B-Instruct | Llama-3.1-70B-Instruct |
| arc_challenge | 0.459 | 0.4838 | 0.5265 | 0.4667 | 0.541 |
| arc_easy | 0.7407 | 0.7765 | 0.7997 | 0.7668 | 0.8026 |
| hellaswag | 0.7043 | 0.7037 | 0.7256 | 0.6913 | 0.7445 |
| mmlu | 0.6043 | 0.5448 | 0.5509 | 0.5312 | 0.5247 |
| openbookqa | 0.36 | 0.394 | 0.416 | 0.388 | 0.424 |
| piqa | 0.7568 | 0.7731 | 0.7731 | 0.7601 | 0.7769 |
| pubmedqa | 0.696 | 0.664 | 0.7 | 0.638 | 0.654 |
| race | 0.4067 | 0.4029 | 0.4364 | 0.3981 | 0.4344 |
| winogrande | 0.6748 | 0.6732 | 0.674 | 0.6606 | 0.6732 |
| truthfulqa | 0.3801 | 0.4202 | 0.4853 | 0.3478 | 0.5028 |
@article{junxiongdaniele2024mambainllama,
title = {The Mamba in the Llama: Distilling and Accelerating Hybrid Models},
author = {Junxiong Wang and Daniele Paliotta and Avner May and Alexander M. Rush and Tri Dao},
journal = {arXiv preprint arXiv:2408.15237},
year = {2024}
}