[RLlib][Data][build_base] Fix RLlib test build rules that should trig…

…ger when there are dataset changes. (ray-project#33683)

* Made the instance type required for the rllib ray data tests larger

Signed-off-by: Kourosh Hakhamaneshi <[email protected]>

.buildkite/pipeline.ml.yml

@@ -141,18 +141,49 @@
       --test_arg=--framework=torch
       rllib/...
 
-- label: ":brain: RLlib: RLlib tests on ray data"
+- label: ":brain: RLlib: Learning tests Pytorch (With Ray Data)"
   conditions: ["NO_WHEELS_REQUIRED", "RAY_CI_DATA_AFFECTED"]
-  instance_size: medium
+  instance_size: large
   commands:
     # skip on master cause we are running these test under all RLlib suites anyw/ay
-    - if [ "$BUILDKITE_PULL_REQUEST" != "false" ]; then exit 0; fi
+    - if [ "$BUILDKITE_PULL_REQUEST" = "false" ]; then exit 0; fi
+    - cleanup() { if [ "${BUILDKITE_PULL_REQUEST}" = "false" ]; then ./ci/build/upload_build_info.sh; fi }; trap cleanup EXIT
+    - RLLIB_TESTING=1 ./ci/env/install-dependencies.sh
+    - ./ci/env/env_info.sh
+    - bazel test --config=ci $(./ci/run/bazel_export_options)
+      --build_tests_only
+      --test_tag_filters=learning_tests_with_ray_data,-multi_gpu,-gpu,-tf_only,-tf2_only
+      --test_arg=--framework=torch
+      rllib/...
+
+
+- label: ":brain: RLlib: Learning tests TF2 (With Ray Data)"
+  conditions: ["NO_WHEELS_REQUIRED", "RAY_CI_DATA_AFFECTED"]
+  instance_size: large
+  commands:
+    # skip on master cause we are running these test under all RLlib suites anyw/ay
+    - if [ "$BUILDKITE_PULL_REQUEST" = "false" ]; then exit 0; fi
+    - cleanup() { if [ "${BUILDKITE_PULL_REQUEST}" = "false" ]; then ./ci/build/upload_build_info.sh; fi }; trap cleanup EXIT
+    - RLLIB_TESTING=1 ./ci/env/install-dependencies.sh
+    - ./ci/env/env_info.sh
+    - bazel test --config=ci $(./ci/run/bazel_export_options)
+      --build_tests_only
+      --test_tag_filters=learning_tests_with_ray_data,-multi_gpu,-gpu,-torch_only
+      --test_arg=--framework=tf2
+      rllib/...
+
+- label: ":brain: RLlib: Unit-tests (With Ray Data)"
+  conditions: ["NO_WHEELS_REQUIRED", "RAY_CI_DATA_AFFECTED"]
+  instance_size: large
+  commands:
+    # skip on master cause we are running these test under all RLlib suites anyw/ay
+    - if [ "$BUILDKITE_PULL_REQUEST" = "false" ]; then exit 0; fi
     - cleanup() { if [ "${BUILDKITE_PULL_REQUEST}" = "false" ]; then ./ci/build/upload_build_info.sh; fi }; trap cleanup EXIT
     - RLLIB_TESTING=1 ./ci/env/install-dependencies.sh
     - ./ci/env/env_info.sh
     - bazel test --config=ci $(./ci/run/bazel_export_options)
       --build_tests_only
-      --test_tag_filters=ray_data,
+      --test_tag_filters=ray_data,-learning_tests_with_ray_data,-multi_gpu,-gpu
       rllib/...
 
 - label: ":brain: RLlib: Algorithm, Model and other tests"

python/ray/data/dataset.py

@@ -209,7 +209,7 @@ class Dataset(Generic[T]):
         >>> ds.map_batches(lambda batch: [v * 2 for v in batch])
         MapBatches(<lambda>)
         +- Dataset(num_blocks=17, num_rows=1000, schema=<class 'int'>)
-        >>> # Compute max.
+        >>> # Compute maximum
         >>> ds.max()
         999
         >>> # Group the data.

rllib/BUILD

@@ -17,6 +17,7 @@
 #    simple CartPole and more advanced variants of it.
 # -- "fake_gpus": Tests that run using 2 fake GPUs.
 # -- "ray_data": Tests that rely on ray_data.
+# -- "learning_tests_with_ray_data": Learning tests that rely on ray_data.
 
 # - Quick algo compilation/tune-train tests, tagged "quick_train".
 #   NOTE: These should be obsoleted in favor of "algorithms_dir" tests as
@@ -247,7 +248,7 @@ py_test(
 py_test(
     name = "learning_tests_pendulum_cql",
     main = "tests/run_regression_tests.py",
-    tags = ["team:rllib", "exclusive", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "ray_data"],
+    tags = ["team:rllib", "exclusive", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "learning_tests_with_ray_data"],
     size = "medium",
     srcs = ["tests/run_regression_tests.py"],
     # Include the zipped json data file as well.
@@ -263,7 +264,7 @@ py_test(
 py_test(
    name = "learning_tests_pendulum_crr",
    main = "tests/run_regression_tests.py",
-   tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "ray_data"],
+   tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "learning_tests_with_ray_data"],
    size = "large",
    srcs = ["tests/run_regression_tests.py"],
    # Include an offline json data file as well.
@@ -277,7 +278,7 @@ py_test(
 py_test(
     name = "learning_tests_cartpole_crr",
     main = "tests/run_regression_tests.py",
-    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "ray_data"],
+    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "learning_tests_with_ray_data"],
     size = "medium",
     srcs = ["tests/run_regression_tests.py"],
     # Include an offline json data file as well.
@@ -291,7 +292,7 @@ py_test(
 py_test(
     name = "learning_tests_cartpole_crr_expectation",
     main = "tests/run_regression_tests.py",
-    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "ray_data"],
+    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "learning_tests_with_ray_data"],
     size = "large",
     srcs = ["tests/run_regression_tests.py"],
     # Include an offline json data file as well.
@@ -392,7 +393,7 @@ py_test(
 py_test(
    name = "learning_tests_pendulum_dt",
    main = "tests/run_regression_tests.py",
-   tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "ray_data"],
+   tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_pendulum", "learning_tests_continuous", "learning_tests_with_ray_data"],
    size = "large",
    srcs = ["tests/run_regression_tests.py"],
    # Include an offline json data file as well.
@@ -406,7 +407,7 @@ py_test(
 py_test(
     name = "learning_tests_cartpole_dt",
     main = "tests/run_regression_tests.py",
-    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "ray_data"],
+    tags = ["team:rllib", "torch_only", "learning_tests", "learning_tests_cartpole", "learning_tests_discrete", "learning_tests_with_ray_data"],
     size = "medium",
     srcs = ["tests/run_regression_tests.py"],
     # Include an offline json data file as well.