Aditya Tomar
Aditya Tomar
Was anyone able to figure out how to get the same results as the paper for this FID benchmark? @z562 any updates?
> you can read in npz file as numpy array and save it as png. This way you have both sets in the same format and can calculate FID Code...
You need to set the `OPENAI_LOGDIR` environment variable to the path where you want the `.pt` file to go: ``` export OPENAI_LOGDIR=/path/to/dir ```
> Anyone Found THE SOLUTION ? There is an alternate method you can use called `GetPointAtDistance()`, which accepts a float as an argument.
Can this feature please be added? It is very annoying to have to manually change the colors of hundreds of runs using the GUI
> 1. At what range of "high" dimensions does this start being better compared to cosine similarity? Does anyone know the answer to this question?
@TitasDas I don't think I have time to investigate this much. I was actually hoping to find an implementation of DIEM so that I can just test it myself and...
> > [@TitasDas](https://github.com/TitasDas) I don't think I have time to investigate this much. I was actually hoping to find an implementation of DIEM so that I can just test it...
@ayshasatti it's training for a long time because it's likely running an infinite loop. Take a look at this [discussion](https://github.com/openai/guided-diffusion/issues/59#issuecomment-1522004239) to help with your question.
@amutesilence check [here](https://github.com/openai/guided-diffusion/issues/59#issuecomment-1522004239)