Karolis Jucys

I can confirm the reward/inventory change mismatch. Attaching a notebook I used to reproduce it using a set seed and hardcoded actions. [test reward vs inventory change by frame.ipynb.zip](https://github.com/minerllabs/minerl/files/6249037/test.reward.vs.inventory.change.by.frame.ipynb.zip)

It should be fine if you comment it out. I didn't work on this part, but based on some slack conversations this happens because "basically there is deadlock if you...

It would be nice, yes, especially if the inventory stays intact on the last observation. Does it?

I noticed that a deterministic policy can give different results on same seed with different realtime_mode settings back in 1.2, so different scores are not surprising. If i remember correctly,...

> fwiw, environments in the experimental directory have no stability guarantees :) I'm trying to add a similar env to biggym and I wanted to use `jump__cheetah` as a template,...

`HumanSurvival` env has the xyz data in either `info` or `obs`, can't remember. There is some commented code here showing how to run it: https://github.com/minerllabs/minerl/blob/dev/tests/basic_test.py There is some other info...

I had both of these issues, first one I solved by doing: `git clone https://github.com/AIcrowd/neurips2020-procgen-starter-kit.git` Second one by changing line 20 of run.sh to: ` export RAY_CPUS=8`

check the info part returned by acting in the HumanSurvival env, it has at least some of the things you're looking for - coordinates, inventory and some more. Some of...

Also remember that the observations are 64x64 in version 0.4, you wouldn't see much detail in the inventory bar or health stats when downsampled so much.

Thanks for the quick response. I did have that suspicion, because the FPS numbers were almost exactly 200.00! Perhaps it would make sense to remove the FPS cap in the...