Add RNN support to torch/PPO and torch/TRPO

[ryanjulian]

[irisliucy]

I'm planning to implement the followings to add RNN support to PPO/ TRPO:

• [x] RNN: lstm_module
• [x] RNN policies and its module for PPO: guassianLSTMModule, guassianLSTMPolicy
• [x] examples (maybe Pendulum): trpo_lstm, ppo_lstm
• [ ] test cases for 1,2

Let me know if I miss anything!

[ryanjulian]

Yes, that sounds good! You should probably pick the more popular of LSTM and GRU.

[nikhilxb]

I'm planning to implement the followings to add RNN support to PPO/ TRPO:

• [x] RNN: lstm_module
• [x] RNN policies and its module for PPO: guassianLSTMModule, guassianLSTMPolicy
• [x] examples (maybe Pendulum): trpo_lstm, ppo_lstm
• [ ] test cases for 1,2

Let me know if I miss anything!

@irisliucy Was this ever completed? I can't seem to find the code for it. I'm trying to implement a recurrent policy in PyTorch to use with PPO and can't figure out the API, so would be helpful to see these examples. Thanks!