taozhiyuai / openbiollm-llama-3-chinese

Advancing Open-source Large Language Models in Medical Domain , in Chinese or mutil-languages.

70B

269 Pulls Updated 5 weeks ago

6 weeks ago

f8e7b3c13e38 · 43GB

model
llama
·
70.6B
·
Q4_K_M
template
{{ if .System }}<|start_header_id|>system<|end_header_id|> {{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|> {{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|> {{ .Response }}<|eot_id|>
params
{"num_keep":24,"stop":["<|start_header_id|>","<|end_header_id|>","<|eot_id|>"]}
system
You are a helpful doctor in China who only reply in Simplified Chinese. Format the reply in MarkDown format.

Readme

import from https://hf-mirror.com/aaditya/Llama3-OpenBioLLM-70B

fJIOPJnY6Ff6fUiSIuMEt.png

Advancing Open-source Large Language Models in Medical Domain

[使用多语言版本] For mutil-language support, visit https://ollama.com/taozhiyuai/openbiollm-llama-3-chinese

[使用英文原版]For original version in English, visit https://ollama.com/taozhiyuai/openbiollm-llama-3

KGmRE5w2sepNtwsEu8t7K.jpeg

Set OpenBioLLM in your language

If you wish OpenBioLLM speak in your language, just set system after running this model by this command. For example, to speak in French:

ollama run taozhiyuai/openbiollm-llama-3-chinese

/set system You are a helpful doctor in France who only reply in French. Format the reply in MarkDown format.

It should work in your language. Strongly suggest 70B which generates better quality texts and follows the system command to speak the language you wish.

Introduction 介绍

生物医学领域优等生,基于LLAMA3打造.

介绍OpenBioLLM-70B：最先进的开源生物医学大型语言模型

OpenBioLLM-70B是专门为生物医学领域设计的高级开源语言模型。该模型由Saama AI Labs开发，利用尖端技术在广泛的生物医学任务中实现最先进的性能。

🏥生物医学专业：OpenBioLLM-70B专为医学和生命科学领域的独特语言和知识要求量身定制。它对大量高质量生物医学数据进行了微调，使其能够以特定领域的准确性和流畅性理解和生成文本。

🎓卓越的性能：OpenBioLLM-70B拥有700亿参数，优于其他类似规模的开源生物医学语言模型。与GPT-4、Gemini、Meditron-70B、Med-PaLM-1和Med-PaLM-2等大型专有和开源模型相比，它还展示了更好的结果。

🧠高级训练技术：OpenBioLLM-70B建立在Meta-Llama-3-70B-Instruct和Meta-Llama-3-70B-Instruct模型的强大基础之上。它结合了DPO数据集和微调配方，以及定制的多样化医疗指导数据集。培训管道的关键组成部分包括：

Screen sample 示例

70B的输出内容多质量高. 设置成为了默认输出中文.如果有时夹杂一些英文,可以让他再次翻译一次,如下图提示词:

70B generate more and quality texts. Chinese is default. Ask the model to translate again, if some English terms mix with texts in your language.

截屏2024-05-08 22.07.00.png

8B的输出内容少,且输出的中文质量不稳定,建议使用中英文版70B.或者8B英文

8B generate less, and 8B Chinese(Mutil-Lan) version is not stable. so 70B is recommended or 8B English version.
截屏2024-05-12 21.10.58.png

截屏2024-05-09 15.21.49.png

Benchemark 医疗模型测评

oPchsJsEpQoGcGXVbh7YS.png
UXF-V0col0Z0sS6BGPBkE.png
截屏2024-05-09 14.25.03.png

WeChat ID: TAOZHIYUAI