diff --git a/tasks/R2R/train.py b/tasks/R2R/train.py
index 17bdac7..a7e2db0 100644
--- a/tasks/R2R/train.py
+++ b/tasks/R2R/train.py
@@ -221,14 +221,15 @@ def train_setup(args, batch_size=BATCH_SIZE):
     return agent, train_env, val_envs
 
 
-def test_setup(args, batch_size=BATCH_SIZE):
-    train_env, test_envs, encoder, decoder = make_env_and_models(
-        args, TRAINVAL_VOCAB, ['train', 'val_seen', 'val_unseen'], ['test'],
-        batch_size=batch_size)
-    agent = Seq2SeqAgent(
-        None, "", encoder, decoder, max_episode_len,
-        max_instruction_length=MAX_INPUT_LENGTH)
-    return agent, train_env, test_envs
+# Test set prediction will be handled separately
+# def test_setup(args, batch_size=BATCH_SIZE):
+#     train_env, test_envs, encoder, decoder = make_env_and_models(
+#         args, TRAINVAL_VOCAB, ['train', 'val_seen', 'val_unseen'], ['test'],
+#         batch_size=batch_size)
+#     agent = Seq2SeqAgent(
+#         None, "", encoder, decoder, max_episode_len,
+#         max_instruction_length=MAX_INPUT_LENGTH)
+#     return agent, train_env, test_envs
 
 
 def train_val(args):
@@ -237,21 +238,22 @@ def train_val(args):
     train(args, train_env, agent, val_envs=val_envs)
 
 
-def test_submission(args):
-    ''' Train on combined training and validation sets, and generate test
-    submission. '''
-    agent, train_env, test_envs = test_setup(args)
-    train(args, train_env, agent)
-
-    test_env = test_envs['test']
-    agent.env = test_env
-
-    agent.results_path = '%s%s_%s_iter_%d.json' % (
-        RESULT_DIR, get_model_prefix(args, train_env.image_features_list),
-        'test', args.n_iters)
-    agent.test(use_dropout=False, feedback='argmax')
-    if not args.no_save:
-        agent.write_results()
+# Test set prediction will be handled separately
+# def test_submission(args):
+#     ''' Train on combined training and validation sets, and generate test
+#     submission. '''
+#     agent, train_env, test_envs = test_setup(args)
+#     train(args, train_env, agent)
+#
+#     test_env = test_envs['test']
+#     agent.env = test_env
+#
+#     agent.results_path = '%s%s_%s_iter_%d.json' % (
+#         RESULT_DIR, get_model_prefix(args, train_env.image_features_list),
+#         'test', args.n_iters)
+#     agent.test(use_dropout=False, feedback='argmax')
+#     if not args.no_save:
+#         agent.write_results()
 
 
 def make_arg_parser():
diff --git a/tasks/R2R/train_speaker.py b/tasks/R2R/train_speaker.py
index f9724f7..ec17ddf 100644
--- a/tasks/R2R/train_speaker.py
+++ b/tasks/R2R/train_speaker.py
@@ -220,13 +220,14 @@ def train_setup(args):
     return agent, train_env, val_envs
 
 
-def test_setup(args):
-    train_env, test_envs, encoder, decoder = make_env_and_models(
-        args, TRAINVAL_VOCAB, ['train', 'val_seen', 'val_unseen'], ['test'])
-    agent = Seq2SeqSpeaker(
-        None, "", encoder, decoder, MAX_INSTRUCTION_LENGTH,
-        max_episode_len=max_episode_len)
-    return agent, train_env, test_envs
+# Test set prediction will be handled separately
+# def test_setup(args):
+#     train_env, test_envs, encoder, decoder = make_env_and_models(
+#         args, TRAINVAL_VOCAB, ['train', 'val_seen', 'val_unseen'], ['test'])
+#     agent = Seq2SeqSpeaker(
+#         None, "", encoder, decoder, MAX_INSTRUCTION_LENGTH,
+#         max_episode_len=max_episode_len)
+#     return agent, train_env, test_envs
 
 
 def train_val(args):
@@ -235,21 +236,22 @@ def train_val(args):
     train(args, train_env, agent, val_envs=val_envs)
 
 
-def test_submission(args):
-    ''' Train on combined training and validation sets, and generate test
-    submission. '''
-    agent, train_env, test_envs = test_setup(args)
-    train(args, train_env, agent)
-
-    test_env = test_envs['test']
-    agent.env = test_env
-
-    agent.results_path = '%s%s_%s_iter_%d.json' % (
-        args.result_dir, get_model_prefix(args, train_env.image_features_list),
-        'test', n_iters)
-    agent.test(use_dropout=False, feedback='argmax')
-    if not args.no_save:
-        agent.write_results()
+# Test set prediction will be handled separately
+# def test_submission(args):
+#     ''' Train on combined training and validation sets, and generate test
+#     submission. '''
+#     agent, train_env, test_envs = test_setup(args)
+#     train(args, train_env, agent)
+#
+#     test_env = test_envs['test']
+#     agent.env = test_env
+#
+#     agent.results_path = '%s%s_%s_iter_%d.json' % (
+#         args.result_dir, get_model_prefix(args, train_env.image_features_list),
+#         'test', n_iters)
+#     agent.test(use_dropout=False, feedback='argmax')
+#     if not args.no_save:
+#         agent.write_results()
 
 
 def make_arg_parser():