MagmaMultiAgent MagMA issues

MagmaMultiAgent / MagMA

Scalable RL Solution for Lux AI S2

MIT License

2 stars 0 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Separate rewards for individual units or groups

#77 GergelyMagyar closed 6 months ago
0
Play tournament with Kaggle models

#76 getlar closed 3 months ago
0
Try different training parameters

#75 getlar closed 3 months ago
0
Play matches with the trained model

#74 getlar closed 3 months ago
0
Train baseline to 2.5M steps

#73 getlar closed 3 months ago
0
Find simplified model that keep factories alive

#72 GergelyMagyar closed 3 months ago
0
Create simplied reward models

#71 GergelyMagyar closed 6 months ago
0
Improve action masking

#70 GergelyMagyar closed 6 months ago
2
Train per unit v.0.0.2 on colab

#69 GergelyMagyar closed 6 months ago
0
Remove unnecessary numpy to pytorch conversions from code

#68 GergelyMagyar closed 6 months ago
0
Refact v0.0.2 original implementation

#67 GergelyMagyar closed 6 months ago
0
Play the trained agent against one of the open source models from Kaggle.

#66 getlar closed 3 months ago
0
Check GPU/CPU/RAM requirements for different num_environment sizes.

#65 getlar closed 3 months ago
0
Train an agent locally and colab.

#64 getlar closed 3 months ago
0
Move Action-per-Agent archiecture over to v0.0.2

#63 GergelyMagyar closed 6 months ago
0
Add action queues for units on v0.0.2

#62 getlar closed 6 months ago
0
Ditch SB3, upgrade to v0.0.2

#61 getlar closed 6 months ago
0
Finish testing results on Wiki for Action per Agent Architecture (v0.0.1)

#60 getlar closed 6 months ago
0
Finish testing results on Wiki for Monolithic Architecture (v0.0.1)

#59 getlar closed 3 months ago
0
Create a way to view the agents during training, without slowing down the fps considerably

#57 GergelyMagyar closed 6 months ago
5
Modify architecture so that actions are generated per unit

#56 GergelyMagyar closed 6 months ago
3
Make sb3 tensor output quick fix vectorized to speed up training

#55 GergelyMagyar closed 10 months ago
0
Make stable baselines handle tensor feature extractor output

#54 GergelyMagyar closed 10 months ago
0
Make factory action model trainable

#53 GergelyMagyar closed 10 months ago
0
Create a feature extractor network and observation parser for factory actions

#52 GergelyMagyar closed 10 months ago
0
Train an agent on the CNN architecture

#51 getlar closed 7 months ago
0
Connect CNN, Obs Parser and Reward Parser together

#50 getlar closed 7 months ago
0
Custom reward parser for CNN

#49 getlar closed 7 months ago
0
Custom M-PPO with Flax

#48 GergelyMagyar closed 7 months ago
0
Update README.md

#47 getlar closed 1 year ago
0
Implement scheduled unit inference on current MPPO model

#46 getlar closed 11 months ago
0
Retrain the CNN on the original observation space encoding

#45 getlar closed 11 months ago
0
Retrain a MPPO model on the extended action and observation space

#44 getlar closed 11 months ago
0
Add factory options to create LIGHT units

#43 getlar closed 11 months ago
0
Add factory actions such as planting lichen

#42 getlar closed 11 months ago
0
Extend the action space of the HEAVY robot to also include self-destruct

#41 getlar closed 11 months ago
0
Redo the conversion of the observation space to layered maps of resources and units.

#40 getlar closed 11 months ago
0
Expand controller to handle multiple robots with the same model

#39 getlar closed 1 year ago
1
Try building 2 HEAVY robots and use them with the single robot model

#38 getlar closed 1 year ago
1
Build a custom CNN network and train an MPPO on CnnPolicy

#37 getlar closed 1 year ago
1
Submit an Agent to Kaggle

#36 getlar closed 1 year ago
1
Evaluate the M-PPO algorithm and log its metrics.

#35 getlar closed 1 year ago
1
Add Stable Baselines 3 Contrib pip package to Docker env.

#34 getlar closed 1 year ago
1
Fix README, add Train and Run Description to the "src" directory.

#33 getlar closed 1 year ago
1
Submit a solution on Kaggle - Stage 1

#32 getlar closed 1 year ago
1
Create a summarized result table for all algorithms, comparing training metrics.

#31 getlar closed 1 year ago
1
Train an agent using Maskable PPO and evaluate it, log metrics.

#30 getlar closed 1 year ago
1
Train an agent using TRPO and evaluate it, log metrics.

#29 getlar closed 1 year ago
1
Train an agent using RecurrentPPO and evaluate it, log metrics.

#28 getlar closed 1 year ago
1
Train an agent using QR-DQN and evaluate it, log metrics.

#27 getlar closed 1 year ago
1