Visualizing my agent's trajectory

#1

Can I use minerl.viewer for visualizing my agent’s behaviour? What format should I use in test.py for saving trajectory to visualize it?

#2

minerl.viewer is for visualizing saved human data. If you would like to visualize your agent’s trajectories simply call env.render. If you are interested in adapting the format, it would be simple enough to translate the env observations to npy and allow them to be loaded by the viewer but at this point that functionality is not implemented. If you make a pull request for this we would be happy to review and add this to MineRL!

Note that malmo has also recently added a recording method for viewing visual observations from the agent - we are currently working on integrating this into the competition!