Aswanth C Manoj

> Check out the Dolphin-llama3 Version that just dropped it fixes many token stop issues for me that were occurring in VScode, they probably fixed other things as well. >...

> > 部署llama3-8b-instruct模型，基于vllm的openai兼容接口。服务拉起来之后，每次请求访问都必须达到最长上下文的输出才会停止。是需要在tokenizer_config.json文件哪里配置停止词吗？ 请求访问如下： { "model": "Meta-Llama-3-8B-Instruct", "messages": [ {"role": "user", "content": "Please tell me a 100 words children's story. Please reply in Chinese."} ], "temperature": 0.1 } 接口返回内容如下： {...

> You should better off training Alpaca format standard from LLaMA-3 pretrained weight with new LLaMa-3 bos/eos token and it should work. Thank you will check it out

> what's the difference between llama3-8b and llama3-8b instruct? if i want to deal with the general text generation task, which one is better? llama3-8B is the base model which...

> I want to labelize json's objects which is better for my task the 8b or 8b instruct? Could you please clarify what you mean by "labelize"? Are you trying...