feat: first approach to action clipping

scripts/hpc/train.pbs

@@ -65,8 +65,8 @@ fi
 
 # TODO Once experiments get serious, change the config
 python scripts/train.py \
-    --env-config-path configs/hpc/smoke_test.yaml \
-    --hyperparameter-config-path configs/hpc/smoke_test.yaml \
+    --env-config-path configs/hpc/debug.yaml \
+    --hyperparameter-config-path configs/hpc/debug.yaml \
     --run-dir "$SCRATCH_RUNDIR"
 
 echo ">>> Staging out results to $DATA_RUNDIR..."

scripts/train.py

@@ -69,18 +69,10 @@ if __name__ == "__main__":
 
     env = make_env(args.env_config_path, args.num_envs)
     raw_env = env.raw
-    print(
-        "\n\n\n Observation space \n",
-        raw_env.observation_space,
-        "Action space \n",
-        raw_env.action_space,
-    )
-    print(
-        "\n\n\n Observation space \n",
-        raw_env.observation_space,
-        "Action space \n",
-        raw_env.action_space,
+    logger.log(
+        {"run_dir": run_dir}
     )
+
     torch.backends.cudnn.deterministic = args.torch_deterministic
 
     ppo_trainer = PPOTrainer(args, env, run_dir, run_name)