Home

Seq2Seq RL Chatbot implementation notes

Links

• Algorithm for Reinforcement Learning

• Log Template

• logs13:Understand Policy Gradient in progress

ideas & concerns

• client id on the chart is broken.
• log policy entropy
• limit iteration to see if the learning works.
• save iterator state

History

• logs0:old
• logs1:Use easy tf log
• logs2:Revisit auth on PyDrive
• logs3:How to save model
• logs4:See if RL framework is working
• logs5:Consolidate logs
• logs6:Make RL model work with small model
• logs7:Test RL with small model
• logs8:Understand policy entropy TO REVISIT
• logs9:Fix msec data issue
• logs10:See if RL works with medium model
• logs11:See if RL works with medium model2
• logs12:Train seq2seq with Large Data
• logs13:Understand Policy Gradient
• logs14:Normalize Reward
• logs15:See if new RL is working
• logs16:RL test shorter reply is better
• logs17:RL test len equals 2 is the best
• logs18:RL test train seq2seq first