Add qmix updates

install_dependencies.sh

@@ -4,6 +4,6 @@
 # conda create -n pymarl python=3.7 -y
 # conda activate pymarl
 
-conda install pytorch torchvision cudatoolkit=9.0 -c pytorch -y
-pip install sacred numpy scipy gym==0.10.8 matplotlib seaborn pyyaml pygame pytest probscale imageio snakeviz tensorboard-logger
+pip install --ignore-installed six
+pip install sacred numpy scipy gym==0.10.8 matplotlib seaborn pyyaml pygame pytest probscale imageio snakeviz tensorboard-logger wandb
 pip install git+https://github.com/oxwhirl/smac.git

install_sc2.sh

@@ -1,18 +1,25 @@
 #!/bin/bash
 # Install SC2 and add the custom maps
 
-pymarl_dir=$HOME/src/pymarl2
-cd $pymarl_dir/3rdparty/
-smac_maps=$pymarl_dir/smac_maps
+if [ -z "$EXP_DIR" ]
+then
+    EXP_DIR=~
+fi
+
+echo "EXP_DIR: $EXP_DIR"
+cd $EXP_DIR/pymarl
+
+mkdir 3rdparty
+cd 3rdparty
 
-export SC2PATH="$HOME/src/pymarl2/3rdparty/StarCraftII"
+export SC2PATH=`pwd`'/StarCraftII'
 echo 'SC2PATH is set to '$SC2PATH
 
 if [ ! -d $SC2PATH ]; then
         echo 'StarCraftII is not installed. Installing now ...';
         wget http://blzdistsc2-a.akamaihd.net/Linux/SC2.4.10.zip
         unzip -P iagreetotheeula SC2.4.10.zip
-        rm -f SC2.4.10.zip
+        rm -rf SC2.4.10.zip
 else
         echo 'StarCraftII is already installed.'
 fi
@@ -28,10 +35,8 @@ fi
 cd ..
 wget https://github.com/oxwhirl/smac/releases/download/v0.1-beta1/SMAC_Maps.zip
 unzip SMAC_Maps.zip
-
-cp -r "$smac_maps/*.SC2Map" ./SMAC_Maps 
 mv SMAC_Maps $MAP_DIR
-rm -f SMAC_Maps.zip
-
+rm -rf SMAC_Maps.zip
 
 echo 'StarCraft II and SMAC are installed.'
+

run.sh

@@ -13,7 +13,7 @@ function onCtrlC () {
 
 config=$1  # qmix
 tag=$2
-maps=${3:-2c_vs_64zg,8m_vs_9m,3s_vs_5z,5m_vs_6m,3s5z_vs_3s6z,corridor,6h_vs_8z,27m_vs_30m}   # MMM2 left out
+maps=${3:-2c_vs_64zg,8m_vs_9m,3s_vs_5z,5m_vs_6m,3s5z_vs_3s6z,corridor,6h_vs_8z,27m_vs_30m,MMM2}   # MMM2 left out
 threads=${4:-8} # 2
 args=${5:-}    # ""
 gpus=${6:-0,1,2,3,4,5,6,7}    # 0,1

run_docker.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 HASH=$(cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w 4 | head -n 1)
-WANDB_API_KEY=$(cat $HOME/.netrc | grep 'password' | awk '{print $2}')
+WANDB_API_KEY=$(cat $WANDB_API_KEY_FILE)
 GPU=$1
 name=${USER}_pymarl_GPU_${GPU}_${HASH}
 

src/config/default.yaml

@@ -19,7 +19,7 @@ buffer_cpu_only: True # If true we won't keep all of the replay buffer in vram
 # --- Logging options ---
 use_tensorboard: False # Log results to tensorboard
 use_wandb: true
-project: "improve_smac"
+project: "smac-v2"
 entity: "benellis3"
 group: "ungrouped"
 save_model: False # Save the models to disk

src/utils/logging.py

@@ -27,9 +27,9 @@ def setup_sacred(self, sacred_run_dict):
         self.use_sacred = True
 
     def setup_wandb(self, config):
-        wandb.init(project=config.project, entity=config.entity, group=config.group, name=config.tag)
+        wandb.init(project=config.project, entity=config.entity, group=config.group, name=config.tag, config=config.__dict__)
         wandb.config = config
-        # setup a custom step metric so that we can track 
+        # setup a custom step metric so that we can track
         # environment steps instead of wandb internal episodes
         wandb.define_metric("train/step")
         wandb.define_metric("train/*", step_metric="train/step")
@@ -85,4 +85,3 @@ def get_logger():
     logger.setLevel('DEBUG')
 
     return logger
-