FanZhang91
FanZhang91
I met the same problem too, Did you solve the problem yet?
Later, I change numpy version to 1.18.0, the problem solved
I met the same problem, did you solved it?@ALei-imotion
用main分支代码做finetune, 用inference_hf代码做预测在解析responce中的output时会报错？而且finetune保存的模型，无法直接用composite_demo的形式直接调用？请问如何解决这两个问题？
> ow is the logs in Win10. My situation is the same as you in Win10
> Did you change anything? I tried in windows 10 with Anaconda and Edge/Chrome, there were no problem. Or please open debug console of browser to check if there werre...
> It is hard to say the reason. Do you make sure all parameters are the same with the downloaded model? I use CALVIN debug dataset. Except for the parameter...
> anyone tried they said? With larger dataset and smaller training steps. I tried 200 demos (10-15s each,10hz) with over 300k steps of training, the model performs terrible, simple pick&place...
@dosu在使用bge-m3编码器+faiss向量存储之后，相对于bge-large-zh-v1.5而言从检索效果来看并没有太大的变化。
> Thanks for posting this. You'll have to elaborate on what exactly you are seeing and how you are running this. Which is the action information you are referring to?...