tidy typos

.gitignore

@@ -1,3 +1,5 @@
+outputs/
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

ddpg/model.py

@@ -121,7 +121,7 @@ def create_act_update_fns(actor, critic, dynamics, target_actor, target_critic,
         next_observations_pred = dynamics(observations, actions)
         dynamics_loss = criterion(
             next_observations_pred,
-            to_tensor(next_obsno proervations),
+            to_tensor(next_observations),
             weights=torch.stack([weights, weights, weights], 1)
         )
         dynamics.zero_grad()

ddpg/train.py

@@ -194,12 +194,12 @@ def train(args, model_fn, act_update_fns, multi_thread, train_single, play_singl
     processes = []
     best_reward = Value("f", 0.0)
 
-    # debugging
-    args.thread = 1
-    multi_thread(actor, critic, dynamics, target_actor, target_critic, target_dynamics, args, act_update_fns, best_reward)
-
     try:
-        if args.num_threads == args.num_train_threads:
+        if args.num_threads == args.num_train_threads == 1:
+            # run a single thread in the foreground so we can debug easier
+            args.thread = 1
+            multi_thread(actor, critic, dynamics, target_actor, target_critic, target_dynamics, args, act_update_fns, best_reward)
+        elif args.num_threads == args.num_train_threads:
             for rank in range(args.num_threads):
                 args.thread = rank
                 p = mp.Process(