weidler/RLaSpa

Issues

• Create template for experiments in Wiki

  #34 opened 6 years ago by HansBambel
  0

• Double Check the Workings of all Networks

  #39 opened 6 years ago by weidler
  0

• Include MultiTask Possibility into Framework

  #29 opened 6 years ago by weidler
  0

• Prepare job script for Aachen cluster

  #33 opened 6 years ago by dannigt
  0

• Create new Tasks

  #28 opened 6 years ago by HansBambel
  1

• Make tensorboard logs get their own directories

  #43 opened 6 years ago by weidler
  0

• Use 4 frames of Race/Evasion as one state

  #42 opened 6 years ago by alessandroscoppio
  0

• Check how DQN memory influence training when using ckpts, eventually save memory in ckpts as well.

  #38 opened 6 years ago by alessandroscoppio
  0

• Add agent.save() and agent.load()

  #40 opened 6 years ago by dannigt
  0

• Create different tasks/obstacle maps to "transfer"

  #17 opened 6 years ago by HansBambel
  3

• Develop a better metric for the progress report

  #37 opened 6 years ago by weidler
  5

• Verify if GPU is indeed utilized

  #32 opened 6 years ago by dannigt
  0

• Exploration strategy

  #31 opened 6 years ago by adrigrillo
  0

• Create statistics of training time and performance

  #19 opened 6 years ago by HansBambel
  2

• Allow the DQN Policy to backpropagate through the representation module

  #30 opened 6 years ago by weidler
  0

• Have rendered testruns automatically saved to a gif file

  #36 opened 6 years ago by weidler
  0

• Make Race/Evasion tasks customizable (kwargs for the register)

  #35 opened 6 years ago by weidler
  0

• Tensorize everything

  #27 opened 6 years ago by HansBambel
  0

• Check if input is as expected and throw meaningful errors.

  #24 opened 6 years ago by weidler
  1

• Document Code, include type hinting!

  #23 opened 6 years ago by weidler
  1

• Look at the heads of the Janus and Cerberus architecture

  #22 opened 6 years ago by HansBambel
  0

• Batch Learning in Representation Learners

  #16 opened 6 years ago by weidler
  0

• Being able to save and continue training a model

  #18 opened 6 years ago by HansBambel
  4

• Make tasks implement the gym interface

  #20 opened 6 years ago by weidler
  1

• Modular Framework for different RL Approaches

  #10 opened 6 years ago by weidler
  4

• Learn ObstaclePathing in latent space

  #6 opened 6 years ago by weidler
  0

• Try to learn ObstaclePathing

  #4 opened 6 years ago by weidler
  6

• Try to learn Pathing in latent space

  #1 opened 6 years ago by weidler
  0

• Test RL agent on (padded) two-task latent representation

  #3 opened 6 years ago by dannigt
  0

• Implement SiameseAutoencoder

  #5 opened 6 years ago by weidler
  1

• Make pathing task use numpy arrays for faster conversions

  #9 opened 6 years ago by HansBambel
  3

• Create DQN framework

  #7 opened 6 years ago by weidler
  2

• Implement

  #8 opened 6 years ago by weidler
  0

• qwdqwd

  #2 opened 6 years ago by weidler
  0