Johan Edstedt

@meshula @404Vector One way to solve this is to use .pyi files which is basically declaring the python interface of your cpython methods. This will make intellisense be able to...

We did it by correlating the gp posterior with coordinate embeddings on the grid. The posterior is unimodal in embedding space, but that space is able to express multimodality in...

You can try fp16, see my code in roma for some examples. I haven't gotten torch.compile to work, but it could improve things.

Hi, the randomness comes both from the network and the ransac. I think the results are usually more even at the end of the training due to reduced learning rate....

Actually you can use any constant you want, as long as you use h,w. You could use 1.3 or 7.5 etc. If you want the model to output exactly in...

@pablospe Currently it will build on push to master, which I think is reasonable. I guess for other CI stuff we want to run on PRs, especially the test suite....

Sorry, I did not mean to cause this review. Was just gonna fix stuff over at my fork. Perhaps we can disable the autoreview? I'll get back to you later.

@pablospe Ok, but the action won't run without the changes to rest in the PR, if I push the action first it will start running on pushes to master. Wouldn't...

@pablospe The issue is that we're building wheels, and the binaries don't end up in any convienent place as far as I can tell. In the basic CI that you...

@pablospe I followed your strategy above (see https://github.com/PoseLib/PoseLib/pull/95) , but this still only runs the already existing action (main.yml). I don't think its possible to run new CI in a...