Antonin RAFFIN
Antonin RAFFIN
> however, does doing flatten make it a discrete space? why? 1D array is just a way of presenting data, as we will reshape it afterward. The main difference between...
> the masking could be dependent on both the current state and masking of other actions. I'm still not sure to get it. The masking of other actions would depend...
then your problem can be define with a Discrete space, not a multi discrete one.
if you install sb3 master version, you must do the same for sb3 contrib. PS: please don't post off topic comments.
Hello, you are right, the callback is actually accessing a global variable defined earlier, it would be cleaner to have it as argument. and thanks for the kind words =)
Hello, this is due to our base VecEnv interface, we need to implement dummy methods for that.
Hello, thanks =) I will give it a try in case the results can be reproduced with different seeds. (it also seem you used twice the budget of the pretrained...
>Which budget are you referring to? I meant the number of timesteps you are using to train the agent: 1M steps for the pretrained agent vs 2M according to your...
Hello, I supposed you are using a cluster? Make sure you did not use more cluster resources (RAM, time, ...) that you asked for...
> I guess we will need to use the normalization during the inference, right? yes, we already save the mean and std for observation in a separate file (`vecnormalize.pkl`) but...