-
Revert the reward function E
-
Improve the model-->resnet 18 Cs
-
Hyperparameter optimization
-
CUDA optimization
-
Batched rollout generation A
-
Lines visualization, periodically M
-
Remove N^4 linear layer Cs
-
Dynamic kernel size Cs
-
Metric, observability, useful metrics to monitor Zs
-
Integrate the ResNet model Cs
-
Rewrite C++ wrapper A
-
Make ConvNet parameterizable
-
Documentation A