Logs24: Run RL on large seq2seq

Log Type	Detail
1: What specific output am I working on right now?	Just run RL on large data to see if it's working.
2: Thinking out loud - hypotheses about the current problem - what to work on next - how can I verify	If reward goes up or replies looks better. It should be working.
3: A record of currently ongoing runs along with a short reminder of what question each run is supposed to answer
4: Results of runs and conclusion	reward was constantly almost the same, which is not expected. Also replies soon became a junk.
5: Next steps	Print some useful information for each step, so that we can catch them.
6: mega.nz

{'machine': 'client1', 'batch_size': 64, 'num_units': 1024, 'num_layers': 3, 'vocab_size': 60000, 'embedding_size': 1024, 'learning_rate': 0.5, 'learning_rate_decay': 0.99, 'use_attention': True, 'encoder_length': 28, 'decoder_length': 28, 'max_gradient_norm': 5.0, 'beam_width': 2, 'num_train_steps': 20000, 'model_path': 'model/tweet_large_rl'}