Maxwell J Jacobson

I see that the results for mujoco are posted. What results did you get for the Atari envs? Was the agent able to more-or-less converge after 1 million steps?

Is there an easy way to get specific data from MetaEvaluator? For example, pulling out the return for each episode in each meta-testing task? This would allow looking at a...

When I use the model for inpainting, there is a large chance that an object will fail to inpaint, and instead GLIDE will simply guess at the background without inserting...