Yingqiang Ge

the continuous version calculates pdf instead of cdf, which is not correct and accurate

> Could you run Dopamine/Recsim using Tensorflow 1.1.5? > https://github.com/google/dopamine/blob/master/README.md#mac-os-x > We will think about a long term solution. Thanks. At first, I couldn't because of the wrong tf version....

> @TobyGE Hi bro, > > I am also using the G5 series as my main instances (G5.xlarge, G5.12xlarge). Everything is fine. > > This is my setup: `git submodule...

> we got a similar problem before. hardwares: > > 1. x86_64 > 2. 4xA10 w/o nvlink > > we change the option and the error disappears: "use_custom_all_reduce": true ->...

I tried Vicuna, got some feedback, a little better. But the actions were not executed. Don't know why.

> > I tried Vicuna, got some feedback, a little better. But the actions were not executed. Don't know why. > > could you show your codes? Similar to @WeixuanXiong...

> > What are the error logs? > > It could be this when using flant5. > > > > As for bloom, it just can't output the format like...

> I dug into it and found the problem. The stop sequences [here](https://github.com/hwchase17/langchain/blob/master/langchain/agents/agent.py#L50) should prevent the LLM from outputting its own hallucinated observations. But the stop sequences for ZeroShotAgent contain...

> which model are you using? From the output, it should be alpaca or even llama.

Have you solved this issue? By changing accelerate version to 0.22.0?