Hongye Liu
Hongye Liu
> Answer: > > 1. If the message is from room, it will say something to the room, if the message is not from room, but contact, it will say...
看起来整个网站都有点bug，左侧所有button都失效了 
对，是的。当时是为了先跑通，直接使用了Embedding，貌似原作者后面有再次进行处理你可以看一下。 ------------------ 原始邮件 ------------------ 发件人: "heyxhh"
Hello, I also interest your dataset, could you share it with me. Thanks! @royxss
> Try changing the pad token ID to 128001 in both the model configuration and the tokenizer. In the pipeline, this is handled automatically to ensure end-to-end generation. It happened...
this situation also happened on llama 70B version.
I eventually found out the solution: In the original code, the following needs to be added: ``` from unsloth.chat_templates import get_chat_template tokenizer = get_chat_template( tokenizer, chat_template = "llama-3.1", mapping =...
Hi daniel, I just tried running the inference with a batch_size>1, but the issue still persists. The template I mentioned above does not work. Instead AutoModelForCausalLM works well. There are...
same issue @eric-haibin-lin
Or any recommend docker image for multi-node? It needs to support agent loop, like retool project. It would be even better if the image has already been tested or run...