LLaMA Pro: Progressive LLaMA with Block Expansion
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929365080604672 |
|---|---|
| author | Wu, Chengyue Gan, Yukang Ge, Yixiao Lu, Zeyu Wang, Jiahao Feng, Ye Shan, Ying Luo, Ping |
| author_facet | Wu, Chengyue Gan, Yukang Ge, Yixiao Lu, Zeyu Wang, Jiahao Feng, Ye Shan, Ying Luo, Ping |
| contents | Humans generally acquire new skills without compromising the old; however, the opposite holds for Large Language Models (LLMs), e.g., from LLaMA to CodeLLaMA. To this end, we propose a new post-pretraining method for LLMs with an expansion of Transformer blocks. We tune the expanded blocks using only new corpus, efficiently and effectively improving the model's knowledge without catastrophic forgetting. In this paper, we experiment on the corpus of code and math, yielding LLaMA Pro-8.3B, a versatile foundation model initialized from LLaMA2-7B, excelling in general tasks, programming, and mathematics. LLaMA Pro and its instruction-following counterpart (LLaMA Pro-Instruct) achieve advanced performance among various benchmarks, demonstrating superiority over existing open models in the LLaMA family and the immense potential of reasoning and addressing diverse tasks as an intelligent agent. Our findings provide valuable insights into integrating natural and programming languages, laying a solid foundation for developing advanced language agents that operate effectively in various environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_02415 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | LLaMA Pro: Progressive LLaMA with Block Expansion Wu, Chengyue Gan, Yukang Ge, Yixiao Lu, Zeyu Wang, Jiahao Feng, Ye Shan, Ying Luo, Ping Computation and Language Humans generally acquire new skills without compromising the old; however, the opposite holds for Large Language Models (LLMs), e.g., from LLaMA to CodeLLaMA. To this end, we propose a new post-pretraining method for LLMs with an expansion of Transformer blocks. We tune the expanded blocks using only new corpus, efficiently and effectively improving the model's knowledge without catastrophic forgetting. In this paper, we experiment on the corpus of code and math, yielding LLaMA Pro-8.3B, a versatile foundation model initialized from LLaMA2-7B, excelling in general tasks, programming, and mathematics. LLaMA Pro and its instruction-following counterpart (LLaMA Pro-Instruct) achieve advanced performance among various benchmarks, demonstrating superiority over existing open models in the LLaMA family and the immense potential of reasoning and addressing diverse tasks as an intelligent agent. Our findings provide valuable insights into integrating natural and programming languages, laying a solid foundation for developing advanced language agents that operate effectively in various environments. |
| title | LLaMA Pro: Progressive LLaMA with Block Expansion |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2401.02415 |