search

hill-a / stable-baselines

A fork of OpenAI Baselines, implementations of reinforcement learning algorithms
http://stable-baselines.readthedocs.io/
MIT License
4.13k stars 724 forks source link

[QUESTION] DQN Policy Kwargs depreciated? #888

Closed ghost closed 4 years ago

ghost commented 4 years ago

SOLVED

Miffyli commented 4 years ago

I am not quite sure what is the problem. DQN does still accept policy_kwargs as before (although it seems to be missing from the docstring, wups), which is then passed on to the policy that is created. There was discussion on the confusion of how dueling architecture works (#587), however, so you have to be careful on what the different parameters will actually do to the network.

araffin commented 4 years ago

my guess is replacing "net_arch" by "layers" should solve your issue ;) (cf dqn doc)

ghost commented 4 years ago

my guess is replacing "net_arch" by "layers" should solve your issue ;) (cf dqn doc)

Yes, but it just doesn't accept the policy_kwargs argument at all, so i can't input net_arch nor layers.

ghost commented 4 years ago

Sorry for semi wasting your time, it got solved by simply reinstalling Stable Baselines. Yes i was using the correct version of SB, somehow reinstalling solved the issue. Don't ask me how.. Sad time.

I'd like to end with that i appreciate the work you put in Stable Baselines, thanks!

ghost commented 3 years ago

Please delete issue