Function train has a Cognitive Complexity of 38 (exceeds 10 allowed). Consider refactoring.

Open

    def train(self):
        clock = self.body.env.clock
        if self.to_train == 1:
            net_util.copy(self.net, self.old_net)  # update old net
            batch = self.sample()

• Read up
  Read up

Function step has a Cognitive Complexity of 26 (exceeds 10 allowed). Consider refactoring.

Open

    def step(self, closure=None):
        loss = None
        if closure is not None:
            loss = closure()

• Read up
  Read up

Function dev_check_train_step has a Cognitive Complexity of 22 (exceeds 10 allowed). Consider refactoring.

Open

def dev_check_train_step(fn):
    '''
    Decorator to check if net.train_step actually updates the network weights properly
    Triggers only if to_check_train_step is True (dev/test mode)
    @example

• Read up
  Read up

Function flatten_dict has a Cognitive Complexity of 18 (exceeds 10 allowed). Consider refactoring.

Open

def flatten_dict(obj, delim='.'):
    '''Missing pydash method to flatten dict'''
    nobj = {}
    for key, val in obj.items():
        if ps.is_dict(val) and not ps.is_empty(val):

• Read up
  Read up

Function log_tensorboard has a Cognitive Complexity of 17 (exceeds 10 allowed). Consider refactoring.

Open

    def log_tensorboard(self):
        '''
        Log summary and useful info to TensorBoard.
        NOTE this logging is comprehensive and memory-intensive, hence it is used in dev mode only
        '''

• Read up
  Read up

Function make_gym_env has a Cognitive Complexity of 16 (exceeds 10 allowed). Consider refactoring.

Open

def make_gym_env(name, seed=None, frame_op=None, frame_op_len=None, image_downsize=None, reward_scale=None, normalize_state=False, episode_life=True):
    '''General method to create any Gym env; auto wraps Atari'''
    env = gym.make(name)
    if seed is not None:
        env.seed(seed)

• Read up
  Read up

Function init_nets has 26 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    def init_nets(self, global_nets=None):
        '''
        Networks: net(actor/policy), q1_net, target_q1_net, q2_net, target_q2_net
        All networks are separate, and have the same hidden layer architectures and optim specs, so tuning is minimal
        '''

Function subproc_worker has a Cognitive Complexity of 13 (exceeds 10 allowed). Consider refactoring.

Open

def subproc_worker(
        pipe, parent_pipe, env_fn_wrapper,
        obs_bufs, obs_shapes, obs_dtypes, keys):
    '''
    Control a single environment instance using IPC and shared memory. Used by ShmemVecEnv.

• Read up
  Read up

Avoid deeply nested control flow statements.

Open

                        for k, v in minibatch.items():
                            if k not in ('advs', 'v_targets'):
                                minibatch[k] = math_util.venv_pack(v, self.body.env.num_envs)
                    advs, v_targets = minibatch['advs'], minibatch['v_targets']

Function plot_trial has a Cognitive Complexity of 12 (exceeds 10 allowed). Consider refactoring.

Open

def plot_trial(trial_spec, trial_metrics, ma=False):
    '''
    Plot the trial graphs:
    - mean_returns, strengths, sample_efficiencies, training_efficiencies, stabilities (with error bar)
    - consistencies (no error bar)

• Read up
  Read up

Function tick has a Cognitive Complexity of 12 (exceeds 10 allowed). Consider refactoring.

Open

def tick(spec, unit):
    '''
    Method to tick lab unit (experiment, trial, session) in meta spec to advance their indices
    Reset lower lab indices to -1 so that they tick to 0
    spec_util.tick(spec, 'session')

• Read up
  Read up

Avoid too many return statements within this function.

Open

        return 'multi_binary'

Function init_params has a Cognitive Complexity of 11 (exceeds 10 allowed). Consider refactoring.

Open

def init_params(module, init_fn):
    '''Initialize module's weights using init_fn, and biases to 0.0'''
    bias_init = 0.0
    classname = util.get_class_name(module)
    if 'Net' in classname:  # skip if it's a net, not pytorch layer

• Read up
  Read up

Function init_global_nets has a Cognitive Complexity of 11 (exceeds 10 allowed). Consider refactoring.

Open

def init_global_nets(algorithm):
    '''
    Initialize global_nets for Hogwild using an identical instance of an algorithm from an isolated Session
    in spec.meta.distributed, specify either:
    - 'shared': global network parameter is shared all the time. In this mode, algorithm local network will be replaced directly by global_net via overriding by identify attribute name

• Read up
  Read up

Function step has a Cognitive Complexity of 11 (exceeds 10 allowed). Consider refactoring.

Open

    def step(self, action):
        obs, reward, done, info = self.env.step(action)
        self.tracked_reward += reward
        # fix shape by inferring from reward
        if np.isscalar(self.total_reward) and not np.isscalar(reward):

• Read up
  Read up

Function plot_multi_trial has a Cognitive Complexity of 11 (exceeds 10 allowed). Consider refactoring.

Open

def plot_multi_trial(trial_metrics_path_list, legend_list, title, graph_prepath, ma=False, name_time_pairs=None, frame_scales=None, palette=None, showlegend=True):
    '''
    Plot multiple trial graphs together
    This method can be used in analysis and also custom plotting by specifying the arguments manually
    @example

• Read up
  Read up