switch to absolute imports

Author: falcondai

methods/__init__.py renamed to rl_baseline/__init__.py

@@ -1,8 +1,9 @@
-from registration import env_registry, method_registry, optimizer_registry
 import torch
 from torch import nn
 import torch.nn.functional as f
-from methods.a2c import A2CModel
+
+from rl_baseline.registration import env_registry, method_registry, optimizer_registry
+from rl_baseline.methods.a2c import A2CModel
 
 # Define your own model. As long as it inherits from the compatible model class, the desired trainer (in this case, A2C) can use it.
 class MyA2CModel(A2CModel):

core.py renamed to rl_baseline/core.py

@@ -7,9 +7,9 @@
 from torch.autograd import Variable
 import torch.nn.functional as f
 
-from core import StochasticPolicy, StateValue
-from registry import method_registry, model_registry, optimizer_registry
-from util import global_norm, log_format, write_tb_event
+from rl_baseline.core import StochasticPolicy, StateValue
+from rl_baseline.registry import method_registry, model_registry, optimizer_registry
+from rl_baseline.util import global_norm, log_format, write_tb_event
 
 
 # Set up logger

examples/a2c_cartpole.py renamed to rl_baseline/examples/a2c_cartpole.py

@@ -8,9 +8,9 @@
 import torch.nn.functional as f
 from torch.nn.utils import clip_grad_norm
 
-from core import StochasticPolicy, StateValue, ActionValue
-from registry import method_registry, model_registry, optimizer_registry
-from util import global_norm, log_format, write_tb_event, linear_schedule, copy_params
+from rl_baseline.core import StochasticPolicy, StateValue, ActionValue
+from rl_baseline.registry import method_registry, model_registry, optimizer_registry
+from rl_baseline.util import global_norm, log_format, write_tb_event, linear_schedule, copy_params
 
 
 # Set up logger

rl_baseline/methods/__init__.py

@@ -1,9 +1,10 @@
 import inspect
-from core import GymEnvSpecWrapper
-from registry import env_registry, optimizer_registry, method_registry, model_registry
 from gym.envs.registration import registry
 from torch import optim
-from methods import a2c, dqn
+
+from rl_baseline.registry import env_registry, optimizer_registry, method_registry, model_registry
+from rl_baseline.core import GymEnvSpecWrapper
+from rl_baseline.methods import a2c, dqn
 
 # Register all envs in Gym
 for spec in registry.all():

methods/a2c.py renamed to rl_baseline/methods/a2c.py

@@ -1,11 +1,12 @@
 def test_a2c_on_cartpole():
     # Train a simple A2C linear model to solve CartPole-v0
-    from methods.a2c import A2CTrainer, A2CLinearModel
     import gym
     import torch
     import numpy as np
     from torch import optim
 
+    from rl_baseline.methods.a2c import A2CTrainer, A2CLinearModel
+
     # Fix seed for replication
     seed = 777
     torch.manual_seed(seed)

methods/dqn.py renamed to rl_baseline/methods/dqn.py

@@ -1,4 +1,4 @@
-from core import GymEnvSpecWrapper
+from rl_baseline.core import GymEnvSpecWrapper
 
 def test_gym_env_spec_wrapper():
     from gym.envs.registration import registry

registration.py renamed to rl_baseline/registration.py

@@ -1,5 +1,5 @@
-from core import GymEnvSpecWrapper
-from registration import env_registry
+from rl_baseline.core import GymEnvSpecWrapper
+from rl_baseline.registration import env_registry
 
 def test_gym_specs_wrapped():
     spec = env_registry['gym.CartPole-v1']

registry.py renamed to rl_baseline/registry.py

@@ -1,4 +1,4 @@
-from registry import Registry
+from rl_baseline.registry import Registry
 
 reg = Registry()
 

tests/test_a2c.py renamed to rl_baseline/tests/test_a2c.py

@@ -1,4 +1,4 @@
-from util import linear_schedule
+from rl_baseline.util import linear_schedule
 
 def test_linear_scheduler():
     assert linear_schedule(0.5, 0, 100, 200, 150) == 0.25, 'In-between value of linear scheduler.'

tests/test_core.py renamed to rl_baseline/tests/test_core.py

@@ -2,6 +2,8 @@
 from __future__ import division
 from __future__ import print_function
 
+from six.moves import xrange
+
 import logging, itertools
 import numpy as np
 
@@ -14,9 +16,8 @@
 import gym
 gym.undo_logger_setup()
 
-from registration import env_registry, optimizer_registry, model_registry, method_registry
-
-from util import log_format, global_norm, get_cartpole_state, set_cartpole_state, copy_params
+from rl_baseline.registration import env_registry, optimizer_registry, model_registry, method_registry
+from rl_baseline.util import log_format, global_norm, get_cartpole_state, set_cartpole_state, copy_params
 
 logging.basicConfig(format=log_format)
 logger = logging.getLogger(__name__)

tests/test_registration.py renamed to rl_baseline/tests/test_registration.py

@@ -0,0 +1,10 @@
+from setuptools import setup, find_packages
+setup(
+    name='rl_baseline',
+    version='0.0.0',
+    packages=find_packages(),
+    author='Falcon Dai',
+    author_email='[email protected]',
+    keywords=['reinforcement learning', 'machine learning', 'pytorch', 'markov decision process'],
+    description='PyTorch implementation of state-of-the-art reinforcement learning algorithms.',
+)