You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The recommerce simulation expects different configuration files for each RL-Agent.
However the API is not capable of differentiating the files:
When a configuration is send to the API, it will be written on disc. The simulation will read this file again. But the API can only write files with one generic name.
One way would be change the behaviour of the API when saving files. But translating the agent into a file name is very tedious and prone to error.
So we need the simulation to handle that.
Suggestion:
When loading a rl configuration file, the simulation will first check if this file exists. If it doesn't it will look for a file called rl_config.json and only if that does not exist, it should error.
find all places where rl configuration files are loaded
check if file exists before loading it
implement fallback (so check for another file named rl_config.json) and load this file instead
change the name the config file is saved in docker_manager.py from q_learning_config.json to rl_config.json has a needs to be fixed comment
test if it works, including docker test
The text was updated successfully, but these errors were encountered:
felix-20
added
the
- Prio 1
Must-Have: High-Priority Bugs, Features requested by EPIC, Big Refactorings
label
Jul 13, 2022
The
recommerce
simulation expects different configuration files for each RL-Agent.However the API is not capable of differentiating the files:
When a configuration is send to the API, it will be written on disc. The simulation will read this file again. But the API can only write files with one generic name.
One way would be change the behaviour of the API when saving files. But translating the agent into a file name is very tedious and prone to error.
So we need the simulation to handle that.
Suggestion:
When loading a rl configuration file, the simulation will first check if this file exists. If it doesn't it will look for a file called
rl_config.json
and only if that does not exist, it should error.rl_config.json
) and load this file insteaddocker_manager.py
fromq_learning_config.json
torl_config.json
has aneeds to be fixed
commentThe text was updated successfully, but these errors were encountered: