David Ding

Hi @JamyIns, Thanks for reporting the issue. Jax has recently deprecated `jax.api`, but older versions of `jaxline` (like the version listed in this project's requirements.txt file) still rely on it....

Hi @dataplayer12, Currently we do not plan to open source the script we used to train the optical flow models, because they depend heavily on our internal infrastructure and unreleased...

Hi @askerlee, We evaluated on the training split for sintel and KITTI, which was also used to generate the results in Table 1 of the [Autoflow paper](https://arxiv.org/pdf/2104.14544.pdf). Note that we...

Hi @aharley, we never tried a smaller batch size. Note that we used a curriculum to train the optical flow model. In addition to the default Autoflow curriculum, we added...

Hi, thanks for your interest in Perceiver IO. We do not plan on open sourcing the training scripts for the masked LM, because the script is heavily tied to our...

Thanks for reporting this issue! This should now be fixed.

Hi @JOBR0, thanks for pointing this out. This is indeed inefficient (but will produce correct results). We will work on a fix for this.

Hi @Enescigdem, Thanks for your interest. Our [colabs](https://github.com/deepmind/deepmind-research/tree/master/perceiver/colabs) contain links to pretrained models and demonstrate how to load and use these models. We have a video autoencoding model that includes...

That's odd. I'll try to get my hands on a Windows machine so I can do more proper testing there.

Here's an example of how to create new windows on the click of a button, and how to close all windows once the main window is closed: ``` import edifice...