MicroRTS Final Tweaks

5125721 Microrts-finetuned-NoWhereToRun uses best model so far
9dcaaed Dummy envs don’t change their size

7bf03d5 Use latest fine-tuned model
554bb9f Add timeout check to RAISocketAI
2c317a7 Test largest map first for performance
160c558 Only ack start once models have been loaded
195b09a (tag: v0.0.24) v0.0.24: RAIPerformanceTournament and fixed pipe RAISocketAI
8a9ff21 RAIPerformanceTournament checks avg turn time quickly

2f7f0f4 Error is not Exception. Catch Throwable instead

827c6dc Fix bots envs never proceding because gameover locked true
fd4fbdb (tag: v0.0.23) Updated MicroRTS Colab demo
bfc93d4 v0.0.23: Pipe interprocess RAISocketAI
fe89107 RAISocketAI uses stdin/stdout pipes instead of socket
5ccabac Fix debug_matrix_obs for microrts_socket_env.py
d5b8345 Even smaller minibatch for BloodBath…
d57def1 Smaller minibatch size for BloodBath
1488696 Training distantresources and bloodbath tweaks
e009199 Auto-lose opponent AI if it throws exception

edb10d2 Shaped reward for NoWhereToRun and DoubleGame
aeb9ccb (tag: v0.0.22) v0.0.22: Python3.11 support and updated models
cf0f2b4 DoubleGame half batch size to fit GPU memory

6f2e7f2 NoWhereToRun and DoubleGame finetuning runs
750014e Output analysis to a csv file
3bcbde6 (tag: v0.0.21) v0.0.21
31abce2 Remove hyperparams that cause a wandb import
2ea3f1e Add a fillna(0) to remove NaN from table
6f94cfd (tag: v0.0.20) v0.0.20 submission
c6977ca Add git-lfs init to setup script
26b3c4a Updated submission models
fd20efc Analyze result of RAIRoundRobinTournament
c887a04 Fix agent support for multiple map sizes
db1c90e Ignore directory structure of unzip to simplify demo