BY571
BY571
Hi @huangwl18, thank you for your quick response! Indeed I failed to install everything directly as listed in the requirements.txt file. Some versions like the python-opencv 4.1.0.25 did work out...
Thanks @jabowery for the feedback, I just added the line to save the behavior function! Regarding your plot... I have no idea. did you plot some other values as well?...
That could possibly be. I haven't tried it though. But if you checked deep mind back then used in their DQN implementation RMSprop as the optimizer. However, since Adam came...
Why I can't converge the network when I use your default setting(terminal: python run_atari_dqn.py)?
You are running PongNoFrameskip-v4? and for how long did you let it train? I have another repository for how to do the plottings: https://github.com/BY571/Tensorboard-Plotting
Why I can't converge the network when I use your default setting(terminal: python run_atari_dqn.py)?
I checked yesterday and noticed the same. for CartPole-v0 and LunarLander-v2 it converges nicely. But as you noticed for Pong it was struggling. I'll have a look later might be...
Why I can't converge the network when I use your default setting(terminal: python run_atari_dqn.py)?
how did your runs work out for cartpole and lunarlander? sry i was quite busy and couldn't dig deeper in the problem yet
> Thanks for this! > > We'll need tests for the feature. > > How do we handle the target parameters? > > Wouldn't something like this be a bit...
Performance with separate target_computation looks good:  But we need to check for speed. It should be similar to our sac implementation.
> @ahmed-touati suggested we use a sampler for this rather than a transform. I'm not strongly opinionated on the matter, mostly because I need more context on what we're trying...
> Why not? I would guess that even if it's a complex nn.Module you can still do pretty much everything with a well tailored function (at least nothing less than...