TianhongDai / esil-hindsight

This is the official code of our paper "Episodic Self-Imitation Learning with Hindsight" [Electronics 2020].
MIT License
7 stars 2 forks source link

hyperparameter #3

Closed quyouyuan closed 3 years ago

quyouyuan commented 3 years ago

Hello! It's convenient for you to answer how many hyperparameters are set when you are running? For example, 'cycles','epoch'. I use the parameters given by readme.md, how can I not run out the results of the model you gave me! Thank you for your answer

TianhongDai commented 3 years ago

@quyouyuan Hi, all of the parameters I used is in: https://github.com/TianhongDai/esil-hindsight/blob/main/arguments.py. I believe if you use the command that I provide in the readme.md, you can definitely reproduce the results. To be noticed, the number of mpi worker should be kept the same as the instruction for the corresponding tasks. For FetchReach task, you can use 1 mpi worker for quick verification. But for other tasks, please use the number of mpi worker provided in the instructions.

Some tasks using different hyperparameters in arguments.py is FetchPickAndPlace and FetchSlide (I also provide this in the readme.md, please check them carefully):