LINT

Signed-off-by: sven1977 <[email protected]>

rllib/algorithms/impala/tests/test_impala_learner.py

@@ -84,9 +84,9 @@ def test_impala_loss(self):
  policy = algo.get_policy()
 
  if fw == "tf2":
- train_batch = SampleBatch(tree.map_structure(
- lambda x: tf.convert_to_tensor(x), FAKE_BATCH
- ))
+ train_batch = SampleBatch(
+ tree.map_structure(lambda x: tf.convert_to_tensor(x), FAKE_BATCH)
+ )
  elif fw == "torch":
  train_batch = convert_to_torch_tensor(SampleBatch(FAKE_BATCH))
 

rllib/algorithms/tests/test_algorithm_config.py

@@ -348,15 +348,19 @@ def get_default_rl_module_spec(self):
  ########################################
  # This is the case where we pass in a multi-agent RLModuleSpec that asks the
  # algorithm to assign a specific type of RLModule class to certain module_ids.
- config = SingleAgentAlgoConfig().rl_module(
- _enable_rl_module_api=True,
- rl_module_spec=MultiAgentRLModuleSpec(
- module_specs={
- "p1": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
- "p2": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
- }
- ),
- ).training(_enable_learner_api=True)
+ config = (
+ SingleAgentAlgoConfig()
+ .rl_module(
+ _enable_rl_module_api=True,
+ rl_module_spec=MultiAgentRLModuleSpec(
+ module_specs={
+ "p1": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
+ "p2": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
+ },
+ ),
+ )
+ .training(_enable_learner_api=True)
+ )
  config.validate()
 
  spec, expected = self._get_expected_marl_spec(config, CustomRLModule1)
@@ -365,10 +369,14 @@ def get_default_rl_module_spec(self):
  ########################################
  # This is the case where we ask the algorithm to assign a specific type of
  # RLModule class to ALL module_ids.
- config = SingleAgentAlgoConfig().rl_module(
- _enable_rl_module_api=True,
- rl_module_spec=SingleAgentRLModuleSpec(module_class=CustomRLModule1),
- ).training(_enable_learner_api=True)
+ config = (
+ SingleAgentAlgoConfig()
+ .rl_module(
+ _enable_rl_module_api=True,
+ rl_module_spec=SingleAgentRLModuleSpec(module_class=CustomRLModule1),
+ )
+ .training(_enable_learner_api=True)
+ )
  config.validate()
 
  spec, expected = self._get_expected_marl_spec(config, CustomRLModule1)
@@ -382,12 +390,16 @@ def get_default_rl_module_spec(self):
  ########################################
  # This is an alternative way to ask the algorithm to assign a specific type of
  # RLModule class to ALL module_ids.
- config = SingleAgentAlgoConfig().rl_module(
- _enable_rl_module_api=True,
- rl_module_spec=MultiAgentRLModuleSpec(
- module_specs=SingleAgentRLModuleSpec(module_class=CustomRLModule1)
- ),
- ).training(_enable_learner_api=True)
+ config = (
+ SingleAgentAlgoConfig()
+ .rl_module(
+ _enable_rl_module_api=True,
+ rl_module_spec=MultiAgentRLModuleSpec(
+ module_specs=SingleAgentRLModuleSpec(module_class=CustomRLModule1)
+ ),
+ )
+ .training(_enable_learner_api=True)
+ )
  config.validate()
 
  spec, expected = self._get_expected_marl_spec(config, CustomRLModule1)
@@ -403,16 +415,20 @@ def get_default_rl_module_spec(self):
  # This is not only assigning a specific type of RLModule class to EACH
  # module_id, but also defining a new custom MultiAgentRLModule class to be used
  # in the multi-agent scenario.
- config = SingleAgentAlgoConfig().rl_module(
- _enable_rl_module_api=True,
- rl_module_spec=MultiAgentRLModuleSpec(
- marl_module_class=CustomMARLModule1,
- module_specs={
- "p1": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
- "p2": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
- },
- ),
- ).training(_enable_learner_api=True)
+ config = (
+ SingleAgentAlgoConfig()
+ .rl_module(
+ _enable_rl_module_api=True,
+ rl_module_spec=MultiAgentRLModuleSpec(
+ marl_module_class=CustomMARLModule1,
+ module_specs={
+ "p1": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
+ "p2": SingleAgentRLModuleSpec(module_class=CustomRLModule1),
+ },
+ ),
+ )
+ .training(_enable_learner_api=True)
+ )
  config.validate()
 
  spec, expected = self._get_expected_marl_spec(
@@ -440,9 +456,11 @@ def get_default_rl_module_spec(self):
  # This is the case where we ask the algorithm to use its default
  # MultiAgentRLModuleSpec, but the MultiAgentRLModuleSpec has not defined its
  # SingleAgentRLmoduleSpecs.
- config = MultiAgentAlgoConfigWithNoSingleAgentSpec().rl_module(
- _enable_rl_module_api=True
- ).training(_enable_learner_api=True)
+ config = (
+ MultiAgentAlgoConfigWithNoSingleAgentSpec()
+ .rl_module(_enable_rl_module_api=True)
+ .training(_enable_learner_api=True)
+ )
 
  self.assertRaisesRegex(
  ValueError,
@@ -454,7 +472,11 @@ def get_default_rl_module_spec(self):
  # This is the case where we ask the algorithm to use its default
  # MultiAgentRLModuleSpec, and the MultiAgentRLModuleSpec has defined its
  # SingleAgentRLmoduleSpecs.
- config = MultiAgentAlgoConfig().rl_module(_enable_rl_module_api=True).training(_enable_learner_api=True)
+ config = (
+ MultiAgentAlgoConfig()
+ .rl_module(_enable_rl_module_api=True)
+ .training(_enable_learner_api=True)
+ )
  config.validate()
 
  spec, expected = self._get_expected_marl_spec(