Virtualhome

This release contains:

- Better control of the character position when talking to objects, in particular ability to fix the position where the character will walk
- Better messages returned by render_script function
- Check when agents perform concurrent invalid actions, such as grabbing the same object
- Bug fixes

The new version of VirtualHome (VH-Social) is out! You can now use VirtualHome to render activities with multiple agents and train policies with those.

Here is what is new.

* VirtualHome now supports multiple agents and a skip_animation mode, to use the environment to train RL models.
* We include OpenAI Gym like environments to train RL agents with VirtualHome.
* API to add characters in the scene and fixed cameras, and record from those cameras

We also have included several bug fixes and performance improvements.

First release