[Bug-fix] RND would not be saved correctly. Added tests (#4514)

ml-agents/mlagents/trainers/tests/torch/saver/test_saver_reward_providers.py

 import pytest
 import os
 
+import numpy as np
+
 from mlagents.trainers.ppo.optimizer_torch import TorchPPOOptimizer
 from mlagents.trainers.sac.optimizer_torch import TorchSACOptimizer
 from mlagents.trainers.model_saver.torch_model_saver import TorchModelSaver
     CuriositySettings,
     GAILSettings,
+    RNDSettings,
+from mlagents.trainers.tests.torch.test_reward_providers.utils import (
+    create_agent_buffer,
+)
 
 DEMO_PATH = (
     os.path.join(os.path.dirname(os.path.abspath(__file__)), os.pardir, os.pardir)
     trainer_settings.reward_signals = {
         RewardSignalType.CURIOSITY: CuriositySettings(),
         RewardSignalType.GAIL: GAILSettings(demo_path=DEMO_PATH),
+        RewardSignalType.RND: RNDSettings(),
     }
     policy = create_policy_mock(trainer_settings, use_discrete=False)
     optimizer = OptimizerClass(policy, trainer_settings)
     module_dict_2 = optimizer2.get_modules()
     assert "Module:GAIL" in module_dict_1
     assert "Module:GAIL" in module_dict_2
+    assert "Module:Curiosity" in module_dict_1
+    assert "Module:Curiosity" in module_dict_2
+    assert "Module:RND-pred" in module_dict_1
+    assert "Module:RND-pred" in module_dict_2
+    assert "Module:RND-target" in module_dict_1
+    assert "Module:RND-target" in module_dict_2
     for name, module1 in module_dict_1.items():
         assert name in module_dict_2
         module2 = module_dict_2[name]
+
+    # Run some rewards
+    data = create_agent_buffer(policy.behavior_spec, 1)
+    for reward_name in optimizer.reward_signals.keys():
+        rp_1 = optimizer.reward_signals[reward_name]
+        rp_2 = optimizer2.reward_signals[reward_name]
+        assert np.array_equal(rp_1.evaluate(data), rp_2.evaluate(data))

ml-agents/mlagents/trainers/torch/components/reward_providers/rnd_reward_provider.py

         self.optimizer.step()
         return {"Losses/RND Loss": loss.detach().cpu().numpy()}
 
+    def get_modules(self):
+        return {
+            f"Module:{self.name}-pred": self._training_network,
+            f"Module:{self.name}-target": self._random_network,
+        }
+
 
 class RNDNetwork(torch.nn.Module):
     EPSILON = 1e-10