Xianglong Tan comments

Results 9 comments of


                                            Xianglong Tan

完全学习不到数据集的内容

猜测 lora 的 rank 不能太大，不然不容易学到特定数据集的知识

Based on the rag project, there are several questions

我可以建一个非官方的微信或者小红书群吗？discordy用不太习惯

windows10下他没用啊！

一样的问题，windows下连VLLM服务都不请求，而且也看不到日志

[Rollout timeout] Loss rollout while training

我可以建一个非官方的微信或者小红书群吗？discord用不太习惯我在小红书上发了篇讨论 agent lightning 的文章，有2K多阅读和500多赞藏，我也希望大家能在国内的平台更方便地进行讨论~

No deployments available for selected model

I get the same problem. However I modified ver/trainer/413 to examine which one of original model path or truncated model path works, neither of them works. So I guess the...

Is is possible to run training task on an agent built by Golang?

I am not sure if this way works: deploy the golang agent as a service, then use python to simply call this service while passing the llm ep provided by...

Training data release

We are annotating complex action sequence data of complex PPT editting task. Is there anyone discussion groups of GUI data annotation?

Is the training on-policy?

Is it possible to make it off-policy?

现在的MLLM真的有视觉Reflection能力吗

这方面预训练数据太少了