internal merge of PR #1256

PiperOrigin-RevId: 223261190

Author: blazejosinski

tensor2tensor/models/research/rl.py

@@ -20,7 +20,6 @@
 import operator
 import gym
 
-from tensor2tensor.data_generators import gym_env
 from tensor2tensor.layers import common_hparams
 from tensor2tensor.layers import common_layers
 from tensor2tensor.layers import discretization

tensor2tensor/rl/trainer_model_free_test.py

@@ -29,7 +29,7 @@
 class TrainTest(tf.test.TestCase):
 
   def test_train_pong(self):
-    hparams = registry.hparams("pong_model_free")
+    hparams = registry.hparams("mfrl_original")
     hparams.batch_size = 2
     hparams.ppo_epochs_num = 2
     hparams.ppo_epoch_length = 3