Sylvain Marchienne issues

Repositories
Issues
Comments

Results 2 issues of


                                            Sylvain Marchienne

Mean instead of sum when computing the `expected_reward` by episode

Hi, According to most of PyTorch REINFORCE algorithm implementations, the policy gradient loss should sum the `log_probs` on the trajectory (sum over t=1...T) instead of computing the mean. In the...

Feature representation of the dataset already computed

Hi, Is there any version of this dataset already preprocessed with frames represented as vectors passed through a CNN model? This would be very convenient to have, for instance like...