Update dependencies (#428)

* Update dependencies

* Fix for numpy v1.17.0

* Downgrade pytest version for travis

* Update .travis.yml

* Rollback to previous docker image

* Update .travis.yml

* Trying to upgrade pytest

* Try to ignore pytest warning

* Update setup.cfg

* Use full name to ignore pytest warning

* Correct import

* Remove gym and tf warnings

* Add TD3 to tensorboard tests

* Ignore additional gym warning

* Get rid of additional tf warning

* Re-enable new docker image

* Test with different tf version

* Upgrade pytest

* Upgrade tf to 1.13.2

* Try downgrading tf

* Move Travis CI test to separate bash file

* Try splitting up test suite

* Diagnostic: echo test glob

* Avoid unintended wildcard

* Upgrade TF to 1.13.2, move scripts to subdirectory

* Split up tests to keep them <20m long

* Try thread-safe SubprocVecEnv

* Disable non-thread safe start methods in tests; document default change

* Rebalance tests

* Rebalance some more

* Rebalance, hopefully for the last time

* Fix globs

* Update docker cpu image: add coverage reporter for travis

* Codacy partial upload

* Bump Docker image version

* Make Travis read environment variable

* Pass project token in

* Remove pip install and fix coverage final report

.travis.yml

@@ -2,23 +2,43 @@ language: python
 python:
   - "3.5"
 
+env:
+  global:
+    - DOCKER_IMAGE=araffin/stable-baselines-cpu:v2.7.1
+
 notifications:
   email: false
 
 services:
   - docker
 
 install:
-  - docker pull araffin/stable-baselines-cpu
+  - docker pull ${DOCKER_IMAGE}
+
+script:
+  - ./scripts/run_tests_travis.sh "${TEST_GLOB}"
 
-matrix:
+jobs:
   include:
-    - name: "Unit Tests"
-      script:
-        # For pull requests from fork, Codacy token is not available, leading to build failure
-        - 'if [ "$TRAVIS_PULL_REQUEST" != "false" ]; then docker run -it --rm --network host --ipc=host --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines-cpu bash -c "cd /root/code/stable-baselines/ && pytest --cov-config .coveragerc --cov-report term --cov=. -v tests/"; fi'
-        - 'if [ "$TRAVIS_PULL_REQUEST" = "false" ]; then docker run -it --env CODACY_PROJECT_TOKEN=$CODACY_PROJECT_TOKEN --rm --network host --ipc=host --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines-cpu bash -c "cd /root/code/stable-baselines/ && pytest --cov-config .coveragerc --cov-report term --cov-report xml --cov=. -v tests/ && python-codacy-coverage -r coverage.xml --token=$CODACY_PROJECT_TOKEN"; fi'
+    # Big test suite. Run in parallel to decrease wall-clock time, and to avoid OOM error from leaks
+    - stage: Test
+      name: "Unit Tests a-h"
+      env: TEST_GLOB="[a-h]*"
+
+    - name: "Unit Tests i-l"
+      env: TEST_GLOB="[i-l]*"
+
+    - name: "Unit Tests m-sa"
+      env: TEST_GLOB="{[m-r]*,sa*}"
+
+    - name: "Unit Tests sb-z"
+      env: TEST_GLOB="{s[b-z]*,[t-z]*}"
 
     - name: "Sphinx Documentation"
       script:
-        - 'docker run -it --rm --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines-cpu bash -c "cd /root/code/stable-baselines/ && pip install .[docs] && pushd docs/ && make clean && make html"'
+        - 'docker run -it --rm --mount src=$(pwd),target=/root/code/stable-baselines,type=bind ${DOCKER_IMAGE} bash -c "cd /root/code/stable-baselines/ && pip install .[docs] && pushd docs/ && make clean && make html"'
+
+    - stage: Codacy Trigger
+      script:
+        # When all test coverage reports have been uploaded, instruct Codacy to start analysis.
+        - 'docker run -it --rm --network host --ipc=host --mount src=$(pwd),target=/root/code/stable-baselines,type=bind --env CODACY_PROJECT_TOKEN=${CODACY_PROJECT_TOKEN} ${DOCKER_IMAGE} bash -c "cd /root/code/stable-baselines/ && /root/code/codacy-coverage-reporter final"'

CONTRIBUTING.md

@@ -91,7 +91,7 @@ Also, when a bug fix is proposed, tests should be added to avoid regression.
 To run tests:
 
 ```
-./run_tests.sh
+./scripts/run_tests.sh
 ```
 
 ## Changelog and Documentation

docker/Dockerfile.cpu

@@ -13,27 +13,27 @@ RUN \
     pip install --upgrade pip && \
     pip install codacy-coverage && \
     pip install scipy && \
-    pip install tqdm && \
     pip install joblib && \
-    pip install zmq && \
-    pip install dill && \
-    pip install progressbar2 && \
     pip install mpi4py && \
     pip install cloudpickle && \
-    pip install tensorflow==1.5.0 && \
-    pip install click && \
+    pip install tensorflow==1.8.0 && \
     pip install opencv-python && \
     pip install numpy && \
     pip install pandas && \
-    pip install pytest==3.5.1 && \
+    pip install pytest && \
     pip install pytest-cov && \
     pip install pytest-env && \
     pip install pytest-xdist && \
     pip install matplotlib && \
-    pip install seaborn && \
-    pip install glob2 && \
     pip install gym[atari,classic_control]>=0.10.9
 
 ENV PATH=$VENV/bin:$PATH
 
+RUN apt-get -y update && apt-get -y install curl jq
+
+# Codacy code coverage report: used for partial code coverage reporting
+RUN cd $CODE_DIR &&\
+    curl -Ls -o codacy-coverage-reporter "$(curl -Ls https://api.github.com/repos/codacy/codacy-coverage-reporter/releases/latest | jq -r '.assets | map({name, browser_download_url} | select(.name | contains("codacy-coverage-reporter-linux"))) | .[0].browser_download_url')" &&\
+    chmod +x codacy-coverage-reporter
+
 CMD /bin/bash

docker/Dockerfile.gpu

@@ -13,25 +13,18 @@ RUN \
     pip install --upgrade pip && \
     pip install codacy-coverage && \
     pip install scipy && \
-    pip install tqdm && \
     pip install joblib && \
-    pip install zmq && \
-    pip install dill && \
-    pip install progressbar2 && \
     pip install mpi4py && \
     pip install cloudpickle && \
     pip install tensorflow-gpu==1.8.0 && \
-    pip install click && \
     pip install opencv-python && \
     pip install numpy && \
     pip install pandas && \
-    pip install pytest==3.5.1 && \
+    pip install pytest && \
     pip install pytest-cov && \
     pip install pytest-env && \
     pip install pytest-xdist && \
     pip install matplotlib && \
-    pip install seaborn && \
-    pip install glob2 && \
     pip install gym[atari,classic_control]>=0.10.9
 
 ENV PATH=$VENV/bin:$PATH

docs/guide/install.rst

@@ -140,7 +140,7 @@ Or, with the shell file:
 
 .. code-block:: bash
 
-   ./run_docker_gpu.sh pytest tests/
+   ./scripts/run_docker_gpu.sh pytest tests/
 
 Run the docker CPU image
 
@@ -152,7 +152,7 @@ Or, with the shell file:
 
 .. code-block:: bash
 
-   ./run_docker_cpu.sh pytest tests/
+   ./scripts/run_docker_cpu.sh pytest tests/
 
 Explanation of the docker command:
 

docs/misc/changelog.rst

@@ -16,6 +16,11 @@ Breaking Changes:
   OpenMPI-dependent algorithms.
   See :ref:`installation notes <openmpi>` and
   `Issue #430 <https://github.com/hill-a/stable-baselines/issues/430>`.
+- SubprocVecEnv now defaults to a thread-safe start method, `forkserver` when
+  available and otherwise `spawn`. This may require application code be
+  wrapped in `if __name__ == '__main__'`. You can restore previous behavior
+  by explicitly setting `start_method = 'fork'`. See
+  `PR #428 <https://github.com/hill-a/stable-baselines/pull/428>`_.
 
 New Features:
 ^^^^^^^^^^^^^
@@ -51,6 +56,34 @@ Documentation:
 - Fix and rename custom policy names (@eavelardev)
 
 
+
+Pre-Release 2.7.1a0 (WIP)
+--------------------------
+
+
+Breaking Changes:
+^^^^^^^^^^^^^^^^^
+- updated dependencies: tensorflow v1.8.0 is now required
+
+New Features:
+^^^^^^^^^^^^^
+
+Bug Fixes:
+^^^^^^^^^^
+- set `allow_pickle=True` for numpy>=1.17.0 when loading expert dataset
+
+Deprecations:
+^^^^^^^^^^^^^
+
+Others:
+^^^^^^^
+- docker images were updated
+
+Documentation:
+^^^^^^^^^^^^^^
+
+
+
 Release 2.7.0 (2019-07-31)
 --------------------------
 

run_docker_cpu.sh → scripts/run_docker_cpu.sh

@@ -8,5 +8,5 @@ echo $cmd_line
 
 
 docker run -it --rm --network host --ipc=host \
- --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines-cpu\
+ --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines-cpu:v2.7.1 \
   bash -c "cd /root/code/stable-baselines/ && $cmd_line"

run_docker_gpu.sh → scripts/run_docker_gpu.sh

@@ -8,5 +8,5 @@ echo $cmd_line
 
 
 docker run -it --runtime=nvidia --rm --network host --ipc=host \
-  --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines\
+  --mount src=$(pwd),target=/root/code/stable-baselines,type=bind araffin/stable-baselines:v2.7.1 \
   bash -c "cd /root/code/stable-baselines/ && $cmd_line"

scripts/run_tests_travis.sh

@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+
+DOCKER_CMD="docker run -it --rm --network host --ipc=host --mount src=$(pwd),target=/root/code/stable-baselines,type=bind"
+BASH_CMD="cd /root/code/stable-baselines/"
+
+if [[ $# -ne 1 ]]; then
+  echo "usage: $0 <test glob>"
+  exit 1
+fi
+
+if [[ ${DOCKER_IMAGE} = "" ]]; then
+  echo "Need DOCKER_IMAGE environment variable to be set."
+  exit 1
+fi
+
+TEST_GLOB=$1
+
+set -e  # exit immediately on any error
+
+# For pull requests from fork, Codacy token is not available, leading to build failure
+if [ "$TRAVIS_PULL_REQUEST" != "false" ]; then
+  ${DOCKER_CMD} ${DOCKER_IMAGE} \
+      bash -c "${BASH_CMD} && \
+               pytest --cov-config .coveragerc --cov-report term --cov=. -v tests/test_${TEST_GLOB}"
+else
+  if [[ ${CODACY_PROJECT_TOKEN} = "" ]]; then
+    echo "Need CODACY_PROJECT_TOKEN environment variable to be set."
+    exit 1
+  fi
+
+  ${DOCKER_CMD} --env CODACY_PROJECT_TOKEN=${CODACY_PROJECT_TOKEN} ${DOCKER_IMAGE} \
+      bash -c "${BASH_CMD} && \
+                pytest --cov-config .coveragerc --cov-report term --cov-report xml --cov=. -v tests/test_${TEST_GLOB} && \
+                /root/code/codacy-coverage-reporter report -l python -r coverage.xml --partial"
+fi

setup.cfg

@@ -4,5 +4,15 @@ license_file = LICENSE
 
 [tool:pytest]
 # Deterministic ordering for tests; useful for pytest-xdist.
-env = 
+env =
 	PYTHONHASHSEED=0
+filterwarnings =
+    ignore:inspect.getargspec:DeprecationWarning:tensorflow
+    ignore::pytest.PytestUnknownMarkWarning
+		# Tensorflow internal warnings
+		ignore:builtin type EagerTensor has no __module__ attribute:DeprecationWarning
+		ignore:The binary mode of fromstring is deprecated:DeprecationWarning
+		ignore::FutureWarning:tensorflow
+		# Gym warnings
+		ignore:Parameters to load are deprecated.:DeprecationWarning
+		ignore:the imp module is deprecated in favour of importlib:PendingDeprecationWarning

setup.py

@@ -12,7 +12,7 @@
 install_tf, tf_gpu = False, False
 try:
     import tensorflow as tf
-    if tf.__version__ < LooseVersion('1.5.0'):
+    if tf.__version__ < LooseVersion('1.8.0'):
         install_tf = True
         # check if a gpu version is needed
         tf_gpu = tf.test.is_gpu_available()
@@ -29,7 +29,7 @@
 
 tf_dependency = []
 if install_tf:
-    tf_dependency = ['tensorflow-gpu>=1.5.0'] if tf_gpu else ['tensorflow>=1.5.0']
+    tf_dependency = ['tensorflow-gpu>=1.8.0'] if tf_gpu else ['tensorflow>=1.8.0']
     if tf_gpu:
         print("A GPU was detected, tensorflow-gpu will be installed")
 
@@ -140,7 +140,7 @@
       license="MIT",
       long_description=long_description,
       long_description_content_type='text/markdown',
-      version="2.7.0",
+      version="2.7.1a0",
       )
 
 # python setup.py sdist

stable_baselines/__init__.py

@@ -7,7 +7,6 @@
 from stable_baselines.td3 import TD3
 from stable_baselines.sac import SAC
 
-
 # Load mpi4py-dependent algorithms only if mpi is installed.
 try:
     import mpi4py
@@ -21,4 +20,4 @@
     from stable_baselines.trpo_mpi import TRPO
 del mpi4py
 
-__version__ = "2.7.0"
+__version__ = "2.7.1a0"

stable_baselines/common/vec_env/subproc_vec_env.py

@@ -55,17 +55,17 @@ class SubprocVecEnv(VecEnv):
     .. warning::
 
         Only 'forkserver' and 'spawn' start methods are thread-safe,
-        which is important when TensorFlow
-        sessions or other non thread-safe libraries are used in the parent (see issue #217).
-        However, compared to 'fork' they incur a small start-up cost and have restrictions on
-        global variables. With those methods,
-        users must wrap the code in an ``if __name__ == "__main__":``
+        which is important when TensorFlow sessions or other non thread-safe
+        libraries are used in the parent (see issue #217). However, compared to
+        'fork' they incur a small start-up cost and have restrictions on
+        global variables. With those methods, users must wrap the code in an
+        ``if __name__ == "__main__":`` block.
         For more information, see the multiprocessing documentation.
 
     :param env_fns: ([Gym Environment]) Environments to run in subprocesses
     :param start_method: (str) method used to start the subprocesses.
            Must be one of the methods returned by multiprocessing.get_all_start_methods().
-           Defaults to 'fork' on available platforms, and 'spawn' otherwise.
+           Defaults to 'forkserver' on available platforms, and 'spawn' otherwise.
     """
 
     def __init__(self, env_fns, start_method=None):
@@ -77,8 +77,8 @@ def __init__(self, env_fns, start_method=None):
             # Fork is not a thread safe method (see issue #217)
             # but is more user friendly (does not require to wrap the code in
             # a `if __name__ == "__main__":`)
-            fork_available = 'fork' in multiprocessing.get_all_start_methods()
-            start_method = 'fork' if fork_available else 'spawn'
+            forkserver_available = 'forkserver' in multiprocessing.get_all_start_methods()
+            start_method = 'forkserver' if forkserver_available else 'spawn'
         ctx = multiprocessing.get_context(start_method)
 
         self.remotes, self.work_remotes = zip(*[ctx.Pipe() for _ in range(n_envs)])

stable_baselines/gail/dataset/dataset.py

@@ -38,7 +38,7 @@ def __init__(self, expert_path=None, traj_data=None, train_fraction=0.7, batch_s
         if traj_data is None and expert_path is None:
             raise ValueError("Must specify one of 'traj_data' or 'expert_path'")
         if traj_data is None:
-            traj_data = np.load(expert_path)
+            traj_data = np.load(expert_path, allow_pickle=True)
 
         if verbose > 0:
             for key, val in traj_data.items():

tests/test_gail.py

@@ -72,7 +72,7 @@ def test_generate(generate_env):
         if key != 'episode_returns':
             assert val.shape[0] == n_timesteps, "inconsistent number of timesteps at '{}'".format(key)
 
-    dataset_loaded = np.load('expert.npz')
+    dataset_loaded = np.load('expert.npz', allow_pickle=True)
     assert dataset.keys() == dataset_loaded.keys()
     for key in dataset.keys():
         assert (dataset[key] == dataset_loaded[key]).all(), "different data at '{}'".format(key)

tests/test_tensorboard.py

@@ -3,7 +3,7 @@
 
 import pytest
 
-from stable_baselines import A2C, ACER, ACKTR, DQN, DDPG, PPO1, PPO2, SAC, TRPO
+from stable_baselines import A2C, ACER, ACKTR, DQN, DDPG, PPO1, PPO2, SAC, TD3, TRPO
 
 TENSORBOARD_DIR = '/tmp/tb_dir/'
 
@@ -19,6 +19,7 @@
     'ppo1': (PPO1, 'CartPole-v1'),
     'ppo2': (PPO2, 'CartPole-v1'),
     'sac': (SAC, 'Pendulum-v0'),
+    'td3': (TD3, 'Pendulum-v0'),
     'trpo': (TRPO, 'CartPole-v1'),
 }
 
@@ -47,5 +48,3 @@ def test_multiple_runs(model_name):
     assert os.path.isdir(TENSORBOARD_DIR + logname + "_1")
     # Check that the log dir name increments correctly
     assert os.path.isdir(TENSORBOARD_DIR + logname + "_2")
-
-

tests/test_vec_envs.py

@@ -265,7 +265,11 @@ def obs_assert(obs):
 
 
 def test_subproc_start_method():
-    start_methods = [None] + multiprocessing.get_all_start_methods()
+    start_methods = [None]
+    # Only test thread-safe methods. Others may deadlock tests! (gh/428)
+    safe_methods = {'forkserver', 'spawn'}
+    available_methods = multiprocessing.get_all_start_methods()
+    start_methods += list(safe_methods.intersection(available_methods))
     space = gym.spaces.Discrete(2)
 
     def obs_assert(obs):