[RLlib; documentation] "RLlib in 60sec" overhaul.

.buildkite/pipeline.yml

@@ -529,7 +529,7 @@
     # "learning_tests|quick_train|examples|tests_dir".
     - bazel test --config=ci $(./scripts/bazel_export_options)
       --build_tests_only
-      --test_tag_filters=-learning_tests,-quick_train,-examples,-tests_dir,-trainers_dir,-flaky,-multi_gpu
+      --test_tag_filters=-learning_tests,-quick_train,-examples,-tests_dir,-trainers_dir,-documentation,-flaky,-multi_gpu
       --test_env=RAY_USE_MULTIPROCESSING_CPU_COUNT=1
       rllib/...
 
@@ -615,6 +615,17 @@
       --test_tag_filters=tests_dir_R,-flaky,-multi_gpu --test_env=RAY_USE_MULTIPROCESSING_CPU_COUNT=1
       rllib/...
 
+- label: ":brain: RLlib: Documentation code/examples"
+  conditions: ["RAY_CI_RLLIB_DIRECTLY_AFFECTED"]
+  commands:
+    - cleanup() { if [ "${BUILDKITE_PULL_REQUEST}" = "false" ]; then ./ci/travis/upload_build_info.sh; fi }; trap cleanup EXIT
+    - RLLIB_TESTING=1 PYTHON=3.7 ./ci/travis/install-dependencies.sh
+    # Because Python version changed, we need to re-install Ray here
+    - rm -rf ./python/ray/thirdparty_files; rm -rf ./python/ray/pickle5_files; ./ci/travis/ci.sh build
+    - bazel test --config=ci $(./scripts/bazel_export_options) --build_tests_only
+      --test_tag_filters=documentation,-flaky --test_env=RAY_USE_MULTIPROCESSING_CPU_COUNT=1
+      rllib/...
+
 - label: ":octopus: Tune tests and examples {1/2; no RLlib}"
   conditions: ["RAY_CI_TUNE_AFFECTED"]
   commands:

doc/source/_static/css/custom.css

@@ -79,4 +79,13 @@ dt:target, span.highlighted {
 }
 img.horizontal-scroll {
     max-width: none;
-}
+}
+
+.clear-both {
+    clear: both;
+}
+
+.buttons-float-left {
+    width: 150px;
+    float: left;
+}

doc/source/index.rst

@@ -318,8 +318,9 @@ Papers
    :maxdepth: -1
    :caption: Ray RLlib
 
-   rllib.rst
+   rllib/index.rst
    rllib-toc.rst
+   rllib/core-concepts.rst
    rllib-training.rst
    rllib-env.rst
    rllib-models.rst

doc/source/rllib-models.rst

@@ -184,6 +184,8 @@ If you would like to provide your own model logic (instead of using RLlib's buil
 can sub-class either ``TFModelV2`` (for TensorFlow) or ``TorchModelV2`` (for PyTorch) and then
 register and specify your sub-class in the config as follows:
 
+.. _tensorflow-models:
+
 Custom TensorFlow Models
 ````````````````````````
 
@@ -240,6 +242,8 @@ More examples and explanations on how to implement custom Tuple/Dict processing
 (also check out `this test case here <https://github.com/ray-project/ray/blob/master/rllib/tests/test_nested_observation_spaces.py>`__),
 custom RNNs, custom model APIs (on top of default models) follow further below.
 
+.. _torch-models:
+
 Custom PyTorch Models
 `````````````````````
 

doc/source/rllib-training.rst

@@ -608,6 +608,8 @@ Custom metrics can be accessed and visualized like any other training result:
 
 .. image:: custom_metric.png
 
+.. _exploration-api:
+
 Customizing Exploration Behavior
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~