基于StarCoder2 7B,擅长编码的非限制版本Dolphin模型家族。

7B

227次拉取 更新于4个月前

说明

DolphinCoder StarCoder2 7b 🐬

latitude.sh 赞助。

加入我们的Discord! https://discord.gg/cognitivecomputations

此模型的开发基于StarCoder2-7b,并遵守bigcode-openrail-m许可。

这只海豚在编码方面非常出色,我使用了大量编码数据进行了训练。

此模型未经审查。我已经过滤了数据集以去除对齐和偏差。这使得模型更符合规范。建议在使用此模型作为服务之前,自己实现对齐层。它将对任何请求,甚至是不道德的请求都高度符合规范。请参阅我关于未经审查的博客文章。 https://erichartford.com/uncensored-models
对于您使用此模型创建的任何内容,您负有责任。请负责任地享受。

训练

使用qLoRA和Axolotl在8x L40S上训练3个epoch花费了2天时间

提示格式
此模型使用ChatML提示格式。

<|im_start|>system
You are DolphinCoder, a helpful AI programming assistant.<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant

示例

<|im_start|>system
You are DolphinCoder, a master at software engineering and coding in any programming language.
<|im_start|>user
Please write me a program in golang that parses all the lines in a file, and reverses them character-wise, and saves it to a new file.
<|im_start|>assistant

量化模型

感谢

  • 本模型的开发得益于 latitude.sh 的慷慨赞助。
  • 欢迎微软加入开源AI!感谢提供Orca-Math数据集!
  • BigCode 的培训 và 公开StarCoder2的权重表示衷心的感谢。
  • 向数据集作者表示衷心的感谢:@ise-uiuc, @teknium, @m-a-p
  • 还要感谢 @winglian 及 Axolotl 团队,为创建最佳训练框架付出了巨大努力!
  • Built with Axolotl
  • 感谢开源人工智能社区中所有教导我、鼓励我的朋友们。

示例输出

如果您想物质支持我的工作

相关周边