GitLab

# Conflicts:
#	definitions.py

An experiment now consists of 5 runs on the test dataset instead of a single run - the mean experiment error is now a more reliable metric of the agent's performance

 Conflicts:
	.gitignore
	README.md
	definitions.py
	requirements.txt
	tests/models/dqn/test_dqn.py
	tests/models/ppo/test_ppo.py

Merge remote-tracking branch 'origin/upload_what_we've_done_so_far' into upload_what_we've_done_so_far

# Conflicts:
#	src/experiments/run.py
#	tests/models/dqn/test_dqn.py

what we've done so far

See merge request !1

Deleted old envs. Implemented the best paper env so far. Added mean epoch reward logging. Changed test run log values to error instead of reward. Added hyperparameter logging.

Changed reward to per step negative absolute error. Added the continuous action space env and an adequate experiment. Plugged the reward logging to Neptune.ai.