readme
Browse files
README.md
CHANGED
|
@@ -1,15 +1,28 @@
|
|
| 1 |
---
|
| 2 |
language:
|
| 3 |
- en
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
---
|
| 5 |
-
|
|
|
|
| 6 |
This repo contains the domain-specific chat model developed from LLaMA-2-Chat-7B, using the method in our paper [Adapting Large Language Models via Reading Comprehension](https://huggingface.co/papers/2309.09530).
|
| 7 |
|
| 8 |
We explore **continued pre-training on domain-specific corpora** for large language models. While this approach enriches LLMs with domain knowledge, it significantly hurts their prompting ability for question answering. Inspired by human learning via reading comprehension, we propose a simple method to **transform large-scale pre-training corpora into reading comprehension texts**, consistently improving prompting performance across tasks in biomedicine, finance, and law domains. **Our 7B model competes with much larger domain-specific models like BloombergGPT-50B**.
|
| 9 |
|
|
|
|
|
|
|
| 10 |
**************************** **Updates** ****************************
|
| 11 |
-
* 12/8: Released our [models](https://huggingface.co/AdaptLLM/finance-chat) developed from LLaMA-2-Chat-7B.
|
| 12 |
-
* 9/18: Released our [paper](https://huggingface.co/papers/2309.09530), [code](https://github.com/microsoft/LMOps), [data](https://huggingface.co/datasets/AdaptLLM/finance-tasks), and [models](https://huggingface.co/AdaptLLM/finance-LLM) developed from LLaMA-1-7B.
|
| 13 |
|
| 14 |
|
| 15 |
## Domain-Specific LLaMA-1
|
|
|
|
| 1 |
---
|
| 2 |
language:
|
| 3 |
- en
|
| 4 |
+
license: llama2
|
| 5 |
+
datasets:
|
| 6 |
+
- EleutherAI/pile
|
| 7 |
+
- Open-Orca/OpenOrca
|
| 8 |
+
- GAIR/lima
|
| 9 |
+
- WizardLM/WizardLM_evol_instruct_V2_196k
|
| 10 |
+
metrics:
|
| 11 |
+
- accuracy
|
| 12 |
+
library_name: fairseq
|
| 13 |
+
pipeline_tag: text-generation
|
| 14 |
---
|
| 15 |
+
|
| 16 |
+
# Adapt (Large) Language Models to Domains
|
| 17 |
This repo contains the domain-specific chat model developed from LLaMA-2-Chat-7B, using the method in our paper [Adapting Large Language Models via Reading Comprehension](https://huggingface.co/papers/2309.09530).
|
| 18 |
|
| 19 |
We explore **continued pre-training on domain-specific corpora** for large language models. While this approach enriches LLMs with domain knowledge, it significantly hurts their prompting ability for question answering. Inspired by human learning via reading comprehension, we propose a simple method to **transform large-scale pre-training corpora into reading comprehension texts**, consistently improving prompting performance across tasks in biomedicine, finance, and law domains. **Our 7B model competes with much larger domain-specific models like BloombergGPT-50B**.
|
| 20 |
|
| 21 |
+
### 🤗 We are currently working hard on developing models across different domains, scales and architectures! Please stay tuned! 🤗
|
| 22 |
+
|
| 23 |
**************************** **Updates** ****************************
|
| 24 |
+
* 12/8: Released our [chat models](https://huggingface.co/AdaptLLM/finance-chat) developed from LLaMA-2-Chat-7B.
|
| 25 |
+
* 9/18: Released our [paper](https://huggingface.co/papers/2309.09530), [code](https://github.com/microsoft/LMOps), [data](https://huggingface.co/datasets/AdaptLLM/finance-tasks), and [base models](https://huggingface.co/AdaptLLM/finance-LLM) developed from LLaMA-1-7B.
|
| 26 |
|
| 27 |
|
| 28 |
## Domain-Specific LLaMA-1
|