Code

Lasertag self play#26

Code

Open

RPegoud wants to merge 57 commits intoRyanNavillus:mainRyanNavillus/Syllabus:mainfrom RPegoud:lasertag_self_playRPegoud/Syllabus:lasertag_self_playCopy head branch name to clipboard

+29,668-534

Commits

Commits on Mar 22, 2024

• added lasertag parallel wrapper

  RPegoud

  committed
  6209199

Commits on Mar 25, 2024

• running lasertag selfplay (not learning)

  RPegoud

  committed
  dd40390

• added opponent update

  RPegoud

  committed
  36c5ea9

• reverted commit

  RPegoud

  committed
  39d4d43

Commits on Mar 26, 2024

• added recording code, run from colab ?

  RPegoud

  committed
  1676dc7

Commits on Mar 28, 2024

• working selfplay lasertag script

  RPegoud

  committed
  f0ea3f9

• updated lasertag script

  RPegoud

  committed
  8134546

• added device selection for selfplay agent storage

  RPegoud

  committed
  1d293d4

• added fsp v0

  RPegoud

  committed
  7abf246

Commits on Mar 29, 2024

• added prioritized fictitious self-play

  RPegoud

  committed
  e7a6ab8

• added win rate and n_games plots for pfsp

  RPegoud

  committed
  990e719

Commits on Apr 2, 2024

• added wandb integration

  RPegoud

  committed
  bcc3d1e

Commits on Apr 3, 2024

• added adversarial sp script

  RPegoud

  committed
  bd12c98

• replaced pettingzoo ppo code with cleanrl base ppo.py

  RPegoud

  committed
  09a34de

Commits on Apr 5, 2024

• Added DualCurriculumWrapper and baselines for SP+DR, FSP+DR, PFSP+DR

  RPegoud

  committed
  e7a1bd3

Commits on Apr 18, 2024

• added dual curriculum wrapper and selfplay variations to syllabus core and curricula

  RPegoud

  committed
  4f42f3a

Commits on Apr 20, 2024

• resolved circular import, integrated PR suggestions, dual curriculum wrapper API changes (still ongoing)

  RPegoud

  committed
  1a5a131

Commits on Apr 22, 2024

• added get_opp and update_agent to base curriculum class

  RPegoud

  committed
  3e43732

• .

  RPegoud

  committed
  7d4a2d8

• Merge branch 'main' of https://github.com/RyanNavillus/Syllabus into lasertag_self_play

  RPegoud

  committed
  82de680

• added pfsp_dr training script with checkpoints, resolved circular import, added Dict task space to dual curriculum

  RPegoud

  committed
  d636b28

Commits on Apr 23, 2024

• added seeding

  RPegoud

  committed
  b79e844

• added single script to run lasertag DR + {SP/FSP/PFSP} and shell script to automate slurm jobs

  RPegoud

  committed
  54f1fc4

Commits on Apr 24, 2024

• added plotly, tqdm, joblib, griddly to setup test dependencies

  RPegoud

  committed
  49ed44b

• refactored typing annotation for python 3.8 compatibility

  RPegoud

  committed
  9131e5c

• fixed tensor.cpu() error

  RPegoud

  committed
  088784c

• corrected wandb logging dir

  RPegoud

  committed
  898c966

• Round robin v0

  RPegoud

  committed
  a44eca0

Commits on Apr 25, 2024

• improved round robin iteration and results logging

  RPegoud

  committed
  d3cd470

Commits on Apr 26, 2024

• fixed pytorch error when running with cuda enabled

  RPegoud

  committed
  1c5b051

• fixed checkpointing

  RPegoud

  committed
  e61a779

Commits on Apr 29, 2024

• adapted training script to run for N updates, added command line arguments, refactoring

  RPegoud

  committed
  cd9599f

• added agent checkpoints based on update steps

  RPegoud

  committed
  b38f37e

Commits on May 1, 2024

• updated round robin result format, added wandb model storage

  RPegoud

  committed
  6b5534d

Commits on May 2, 2024

• PLR + selfplay v0

  RPegoud

  committed
  a35ebcb

Commits on May 3, 2024

• fixed wandb model logging errors

  RPegoud

  committed
  3c30ccf

Commits on May 4, 2024

• added seed to selfplay checkpoints

  RPegoud

  committed
  552dbfc

Commits on May 6, 2024

• added round robin results parsing script

  RPegoud

  committed
  226c491

Commits on May 7, 2024

• Add chaecks for missing files

  RyanNavillus

  committed
  24beb7f

• Fix num_steps

  RyanNavillus

  committed
  69cc927

• lasertag vecenv attempt

  RPegoud

  committed
  f81db5b

• Merge branch 'lasertag_self_play' of https://github.com/RPegoud/Syllabus into lasertag_self_play

  RPegoud

  committed
  feebb4b

Commits on May 13, 2024

• lasertag CNN/LSTM v0

  RPegoud

  committed
  4d0938c

Commits on May 20, 2024

• lstm fixes, selfplay changes (+ to cpu before save)

  RPegoud

  committed
  a4cea85

• rnn lstm fixes, selfplay => agent to gpu after save

  RPegoud

  committed
  348453f

Commits on May 21, 2024

• improved round-robin parsing

  RPegoud

  committed
  ae74fa5

• corrected round robin script for lstm

  RPegoud

  committed
  b991f8b

Commits on Jun 11, 2024

• vecenv progress

  RPegoud

  committed
  ba3349c

Commits on Jul 22, 2024

• functional vecenv / LSTM code, next-steps: add curriculum wrapper

  RPegoud

  committed
  ef35582

Commits on Jul 24, 2024

• Curriculum setup, requires vectorization of env curriculum

  RPegoud

  committed
  9dfaf8d

• added comments

  RPegoud

  committed
  24799ac

• Update to use syllabus infra

  RyanNavillus

  committed
  a8a3ec5

• Merge branch 'main' of https://github.com/RyanNavillus/Syllabus into lasertag_self_play

  RyanNavillus

  committed
  e7070f0

• Fix bad merge

  RyanNavillus

  committed
  6c0ad7c

• Add syllabus infrastructure and batched sampling to agent curriculum

  RyanNavillus

  committed
  df762f8

• Set workers back to 8

  RyanNavillus

  committed
  0bc7a04

Commits on Aug 22, 2024

• working vectorized agent selection with queue storage

  RPegoud

  committed
  035c6dc