Merge branch 'master' of https://github.com/alex-petrenko/sample-factory

.gitignore

@@ -139,3 +139,4 @@ trained_policies/**
 *~
 
 cmake-build*
+envs/doom/multiplayer/tests/_vizdoom/

.travis.yml

@@ -0,0 +1,28 @@
+language: python
+dist: bionic
+python:
+  - "3.7"
+
+notifications:
+  email: false
+
+before_install:
+  - sudo apt-get update
+  - sudo apt-get install libboost-all-dev libsdl2-dev zlib1g-dev libbz2-dev
+  - sudo apt-get install libjpeg-dev libfluidsynth-dev libgme-dev libopenal-dev libmpg123-dev
+  - sudo apt-get install libsndfile1-dev libwildmidi-dev libgtk-3-dev timidity nasm tar chrpath
+
+install:
+  - wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
+  - bash miniconda.sh -b -p $HOME/miniconda
+  - source "$HOME/miniconda/etc/profile.d/conda.sh"
+  - hash -r
+  - conda config --set always_yes yes --set changeps1 no
+  - conda update -q conda
+  - conda info -a
+  - patch environment.yml ./.travis/patchfile.patch
+  - conda env create -q -f environment.yml
+  - conda activate sample-factory
+
+script:
+  - python -m unittest

.travis/patchfile.patch

@@ -0,0 +1,31 @@
+7a8
+>   - astroid=2.4.0=py37_0
+13c14
+<   - cudatoolkit=10.1.243=h6bb024c_0
+---
+>   - cpuonly=1.0=0
+30a32
+>   - isort=4.3.21=py37_0
+33a36
+>   - lazy-object-proxy=1.4.3=py37h7b6447c_0
+50a54
+>   - mccabe=0.6.1=py37_1
+68a73
+>   - pylint=2.5.0=py37_1
+71c76
+<   - pytorch=1.4.0=py3.7_cuda10.1.243_cudnn7.6.3_0
+---
+>   - pytorch=1.5.0=py3.7_cpu_0
+78c83,85
+<   - torchvision=0.5.0=py37_cu101
+---
+>   - toml=0.10.0=py37h28b3542_0
+>   - torchvision=0.6.0=py37_cpu
+>   - typed-ast=1.4.1=py37h7b6447c_0
+109a117
+>     - gym-minigrid==1.0.1
+125a134,135
+>     - tensorflow-cpu==1.15.0
+>     - tensorflow-estimator==1.15.1
+129a140
+>     - git+https://github.com/alex-petrenko/ViZDoom@doom_bot_project#egg=vizdoom

README.md

@@ -1,3 +1,5 @@
+[![Build Status](https://travis-ci.org/alex-petrenko/sample-factory.svg?branch=master)](https://travis-ci.org/github/alex-petrenko/sample-factory)
+
 # Sample Factory
 
 Codebase for high throughput asynchronous reinforcement learning.

algorithms/appo/tests/test_model.py

@@ -1,3 +1,4 @@
+import unittest
 from unittest import TestCase
 
 import torch
@@ -11,7 +12,9 @@
 
 
 class TestModel(TestCase):
-    def test_forward_pass(self):
+
+    @staticmethod
+    def forward_pass(device_type):
         env_name = 'atari_breakout'
         cfg = default_cfg(algo='APPO', env=env_name)
         cfg.actor_critic_share_weights = True
@@ -24,7 +27,7 @@ def test_forward_pass(self):
         torch.backends.cudnn.benchmark = True
 
         actor_critic = create_actor_critic(cfg, env.observation_space, env.action_space)
-        device = torch.device('cuda')
+        device = torch.device(device_type)
         actor_critic.to(device)
 
         timing = Timing()
@@ -43,3 +46,10 @@ def test_forward_pass(self):
                 log.debug('Progress %d/%d', i, n)
 
         log.debug('Timing: %s', timing)
+
+    def test_forward_pass_cpu(self):
+        self.forward_pass('cpu')
+
+    @unittest.skipUnless(torch.cuda.is_available(), 'This test requires a GPU')
+    def test_forward_pass_gpu(self):
+        self.forward_pass('cuda')

envs/doom/multiplayer/doom_multiagent_wrapper.py

@@ -255,7 +255,7 @@ def _ensure_initialized(self):
                             time.sleep(0.05)
 
                     for i, worker in enumerate(self.workers):
-                        worker.result_queue.get(timeout=5)
+                        worker.result_queue.get(timeout=20)
                         worker.result_queue.task_done()
                         worker.task_queue.join()
             except filelock.Timeout: