Release v0.3.0

.github/cloud_builder/run_command_on_active_checkout.yaml

@@ -3,7 +3,7 @@ substitutions:
 options:
   logging: CLOUD_LOGGING_ONLY
 steps:
-  - name: us-central1-docker.pkg.dev/external-snap-ci-github-gigl/gigl-base-images/gigl-builder:7d3182eeb6446ce3e35910babba990c8e003879d.109.1
+  - name: us-central1-docker.pkg.dev/external-snap-ci-github-gigl/gigl-base-images/gigl-builder:6db83bdc98b2da65ac80243ce08cd8b37b3ee85f.110.1
     entrypoint: /bin/bash
     # Route sbt through Google's Maven Central mirror to avoid 429 rate limits from repo1.maven.org.
     # Intentionally set here (CI env) rather than in scala/.sbtopts or scala_spark35/.sbtopts to avoid

CHANGELOG.md

@@ -6,6 +6,90 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
 ## [Unreleased]
 
+## [0.3.0] - Jun 1, 2026
+
+### Deprecations
+
+- Deprecate `ShardStrategy` and default distributed sharding to `CONTIGUOUS` by @kmontemayor2-sc in
+  https://github.com/Snapchat/GiGL/pull/582 and https://github.com/Snapchat/GiGL/pull/545
+- Migrate `RESOURCE_CONFIG_PATH` to `GIGL_RESOURCE_CONFIG_URI` by @kmontemayor2-sc in
+  https://github.com/Snapchat/GiGL/pull/648
+
+### Changed
+
+- Replace mypy with [ty](https://github.com/astral-sh/ty) for static type checking and migrate formatting to Ruff by
+  @svij-sc in https://github.com/Snapchat/GiGL/pull/585 and https://github.com/Snapchat/GiGL/pull/583
+- Consolidate distributed loader and sampler abstractions, including `BaseDistLoader`, `BaseDistNeighborSampler`,
+  sampler factory helpers, two-phase loader initialization, and shared sampling options by @mkolodner-sc and
+  @kmontemayor2-sc in https://github.com/Snapchat/GiGL/pull/495, https://github.com/Snapchat/GiGL/pull/532,
+  https://github.com/Snapchat/GiGL/pull/536, https://github.com/Snapchat/GiGL/pull/570,
+  https://github.com/Snapchat/GiGL/pull/576, https://github.com/Snapchat/GiGL/pull/579, and
+  https://github.com/Snapchat/GiGL/pull/561
+- Merge `testing/` into `tests/`, migrate tests to GiGL test case utilities, and improve graph store integration test
+  coverage by @kmontemayor2-sc and @svij-sc in https://github.com/Snapchat/GiGL/pull/494,
+  https://github.com/Snapchat/GiGL/pull/479, https://github.com/Snapchat/GiGL/pull/480,
+  https://github.com/Snapchat/GiGL/pull/515, and https://github.com/Snapchat/GiGL/pull/547
+
+### Added
+
+- Enable GraphStore mode across storage and compute, including GiGL-owned `DistServer`, `DistABLPLoader` GraphStore
+  mode, multiple GraphStore loaders, and homogeneous and heterogeneous examples by @kmontemayor2-sc in
+  https://github.com/Snapchat/GiGL/pull/476, https://github.com/Snapchat/GiGL/pull/485,
+  https://github.com/Snapchat/GiGL/pull/493, https://github.com/Snapchat/GiGL/pull/514, and
+  https://github.com/Snapchat/GiGL/pull/526
+- Add distributed and C++-based PPR sampling, including PPR sequence generation and new GiGL wheel builds with the
+  `gigl-core` C++/CUDA extension package, by @mkolodner-sc and @yliu2-sc in https://github.com/Snapchat/GiGL/pull/538,
+  https://github.com/Snapchat/GiGL/pull/560, https://github.com/Snapchat/GiGL/pull/558, and
+  https://github.com/Snapchat/GiGL/pull/556
+- Add shared multi-channel graph store sampling backend, remote channels with pinned-memory bulk transfer, and two-phase
+  sampling APIs by @kmontemayor2-sc in https://github.com/Snapchat/GiGL/pull/577,
+  https://github.com/Snapchat/GiGL/pull/565, and https://github.com/Snapchat/GiGL/pull/578
+- Add weighted sampling, positional encoding transforms, Graph Transformer encoder, degree tensor computation for
+  `DistDataset`, and max-label-per-anchor support in the data splitter by @mkolodner-sc and @yliu2-sc in
+  https://github.com/Snapchat/GiGL/pull/635, https://github.com/Snapchat/GiGL/pull/509,
+  https://github.com/Snapchat/GiGL/pull/537, https://github.com/Snapchat/GiGL/pull/517, and
+  https://github.com/Snapchat/GiGL/pull/589
+- Add `CustomResourceConfig` shell-command launchers, custom launcher subprocess dispatch, and GiGL env var propagation
+  for custom and Vertex AI launchers by @kmontemayor2-sc in https://github.com/Snapchat/GiGL/pull/625,
+  https://github.com/Snapchat/GiGL/pull/626, https://github.com/Snapchat/GiGL/pull/642, and
+  https://github.com/Snapchat/GiGL/pull/653
+- Add Vertex AI boot disk, reservation, and Data Preprocessor timeout controls by @zfan3-sc, @kmontemayor2-sc, and
+  @mkolodner-sc in https://github.com/Snapchat/GiGL/pull/521, https://github.com/Snapchat/GiGL/pull/590, and
+  https://github.com/Snapchat/GiGL/pull/524
+- Add GBML config wrapper maps for node and edge type metadata, unified metadata extraction, BigQuery latest-table
+  utility, and SNC example code by @svij-sc, @mkolodner-sc, and @kmontemayor2-sc in
+  https://github.com/Snapchat/GiGL/pull/643, https://github.com/Snapchat/GiGL/pull/544,
+  https://github.com/Snapchat/GiGL/pull/516, and https://github.com/Snapchat/GiGL/pull/641
+
+### Fixed
+
+- Fix PPR sampler output edges and memory behavior by @mkolodner-sc in https://github.com/Snapchat/GiGL/pull/562,
+  https://github.com/Snapchat/GiGL/pull/566, and https://github.com/Snapchat/GiGL/pull/645
+- Fix dataloading of multiple labels, missing anchor-node labels, dataset factory parallel tensor loading, test dataset
+  edge direction, and `PreprocessedMetadataPbWrapper` `LocalUri` kwargs by @mkolodner-sc, @kmontemayor2-sc, and @svij-sc
+  in https://github.com/Snapchat/GiGL/pull/612, https://github.com/Snapchat/GiGL/pull/571,
+  https://github.com/Snapchat/GiGL/pull/606, https://github.com/Snapchat/GiGL/pull/552, and
+  https://github.com/Snapchat/GiGL/pull/639
+- Fix launcher image selection for graph store storage pools, default `should_use_glt_backend`, C++ installation, SBT
+  dependency resolution, and types-protobuf v7 `ParseDict` compatibility by @kmontemayor2-sc, @mkolodner-sc, and
+  @svij-sc in https://github.com/Snapchat/GiGL/pull/615, https://github.com/Snapchat/GiGL/pull/609,
+  https://github.com/Snapchat/GiGL/pull/619, https://github.com/Snapchat/GiGL/pull/631, and
+  https://github.com/Snapchat/GiGL/pull/638
+- Make metric exporters fall back to NoOp when initialization fails and fix seeding utility behavior by @mkolodner-sc
+  and @svij-sc in https://github.com/Snapchat/GiGL/pull/550 and https://github.com/Snapchat/GiGL/pull/644
+
+### Misc
+
+- Add and refine agent/development guidance, including `CLAUDE.md`, `AGENTS.md`, `/all_test`, `/codex-review`, and
+  `/watch-action` workflow support by @kmontemayor2-sc in https://github.com/Snapchat/GiGL/pull/498,
+  https://github.com/Snapchat/GiGL/pull/504, https://github.com/Snapchat/GiGL/pull/543,
+  https://github.com/Snapchat/GiGL/pull/508, https://github.com/Snapchat/GiGL/pull/528, and
+  https://github.com/Snapchat/GiGL/pull/563
+- Update graph store and in-memory SGS documentation, supervision-edge direction docs, Vertex AI Agent Platform links,
+  and presubmit whitespace checks by @mkolodner-sc and @kmontemayor2-sc in https://github.com/Snapchat/GiGL/pull/553,
+  https://github.com/Snapchat/GiGL/pull/593, https://github.com/Snapchat/GiGL/pull/595,
+  https://github.com/Snapchat/GiGL/pull/608, and https://github.com/Snapchat/GiGL/pull/607
+
 ## [0.2.0] - Jan 30, 2025
 
 ### Added

gigl-core/pyproject.toml

@@ -2,7 +2,7 @@
 name            = "gigl-core"
 description     = "GiGL C++/CUDA kernels (pybind11 extensions)"
 readme          = "README.md"
-version         = "0.2.0"
+version         = "0.3.0"
 requires-python = "==3.11.*"
 # Torch is resolved from the ambient environment. gigl-core wheels are ABI-bound
 # to the torch variant they were built against (cpu or cu128). The parent `gigl`

gigl/__init__.py

@@ -1 +1 @@
-__version__ = "0.2.0"
+__version__ = "0.3.0"

gigl/dep_vars.env

@@ -1,13 +1,13 @@
 # Note this file only supports static key value pairs so it can be loaded by make, bash, python, and sbt without any additional parsing.
-DOCKER_LATEST_BASE_CUDA_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-cuda-base:7d3182eeb6446ce3e35910babba990c8e003879d.109.1
-DOCKER_LATEST_BASE_CPU_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-cpu-base:7d3182eeb6446ce3e35910babba990c8e003879d.109.1
-DOCKER_LATEST_BASE_DATAFLOW_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-dataflow-base:7d3182eeb6446ce3e35910babba990c8e003879d.109.1
+DOCKER_LATEST_BASE_CUDA_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-cuda-base:6db83bdc98b2da65ac80243ce08cd8b37b3ee85f.110.1
+DOCKER_LATEST_BASE_CPU_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-cpu-base:6db83bdc98b2da65ac80243ce08cd8b37b3ee85f.110.1
+DOCKER_LATEST_BASE_DATAFLOW_IMAGE_NAME_WITH_TAG=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-dataflow-base:6db83bdc98b2da65ac80243ce08cd8b37b3ee85f.110.1
 
-DEFAULT_GIGL_RELEASE_SRC_IMAGE_CUDA=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cuda:0.2.0
-DEFAULT_GIGL_RELEASE_SRC_IMAGE_CPU=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cpu:0.2.0
-DEFAULT_GIGL_RELEASE_SRC_IMAGE_DATAFLOW_CPU=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cpu-dataflow:0.2.0
-DEFAULT_GIGL_RELEASE_DEV_WORKBENCH_IMAGE=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-dev-workbench:0.2.0
-DEFAULT_GIGL_RELEASE_KFP_PIPELINE_PATH=gs://public-gigl/releases/pipelines/gigl-pipeline-0.2.0.yaml
+DEFAULT_GIGL_RELEASE_SRC_IMAGE_CUDA=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cuda:0.3.0
+DEFAULT_GIGL_RELEASE_SRC_IMAGE_CPU=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cpu:0.3.0
+DEFAULT_GIGL_RELEASE_SRC_IMAGE_DATAFLOW_CPU=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/src-cpu-dataflow:0.3.0
+DEFAULT_GIGL_RELEASE_DEV_WORKBENCH_IMAGE=us-central1-docker.pkg.dev/external-snap-ci-github-gigl/public-gigl/gigl-dev-workbench:0.3.0
+DEFAULT_GIGL_RELEASE_KFP_PIPELINE_PATH=gs://public-gigl/releases/pipelines/gigl-pipeline-0.3.0.yaml
 
 SPARK_31_TFRECORD_JAR_GCS_PATH=gs://public-gigl/tools/scala/spark_packages/spark-custom-tfrecord_2.12-0.5.0.jar
 SPARK_35_TFRECORD_JAR_GCS_PATH=gs://public-gigl/tools/scala/spark_packages/spark_3.5.0-custom-tfrecord_2.12-0.6.1.jar

pyproject.toml

@@ -2,7 +2,7 @@
 name            = "gigl"
 description     = "GIgantic Graph Learning Library"
 readme          = "README.md"
-version         = "0.2.0"
+version         = "0.3.0"
 classifiers     = [
     "Programming Language :: Python",
     "Programming Language :: Python :: 3",
@@ -15,7 +15,7 @@ dependencies = [
   "chardet",
   # gigl-core hosts all C++ / CUDA / pybind11 extensions. Separate wheel per torch
   # variant (cpu/cu128). Version must match gigl exactly.
-  "gigl-core==0.2.0",
+  "gigl-core==0.3.0",
   "google-cloud-aiplatform",
   "google-cloud-dataproc",
   "google-cloud-logging",

scripts/bump_version.py

@@ -99,7 +99,9 @@ def update_pyproject(version: str) -> None:
     path = f"{GIGL_ROOT_DIR}/pyproject.toml"
     with open(path, "r") as f:
         content = f.read()
-    content = re.sub(r'(version\s*)=\s*"[\d\.]+"', f'\\1= "{version}"', content)
+    content = re.sub(
+        r'(?m)^(version[ \t]*)=[ \t]*"[^"]+"', f'\\1= "{version}"', content
+    )
     # Keep the gigl-core pin in sync with the new version.
     content = re.sub(r'"gigl-core==[\d\.a-zA-Z]+"', f'"gigl-core=={version}"', content)
     with open(path, "w") as f:
@@ -110,7 +112,9 @@ def update_gigl_core_pyproject(version: str) -> None:
     path = f"{GIGL_ROOT_DIR}/gigl-core/pyproject.toml"
     with open(path, "r") as f:
         content = f.read()
-    content = re.sub(r'(version\s*)=\s*"[\d\.]+"', f'\\1= "{version}"', content)
+    content = re.sub(
+        r'(?m)^(version[ \t]*)=[ \t]*"[^"]+"', f'\\1= "{version}"', content
+    )
     with open(path, "w") as f:
         f.write(content)
 

tests/integration/distributed/utils/networking_test.py

@@ -1,11 +1,15 @@
 import uuid
 from textwrap import dedent
 
+from google.cloud.aiplatform_v1.types import env_var
 from parameterized import param, parameterized
 
 from gigl.common.constants import DEFAULT_GIGL_RELEASE_SRC_IMAGE_CPU
 from gigl.common.services.vertex_ai import VertexAiJobConfig, VertexAIService
+from gigl.common.utils.proto_utils import ProtoUtils
+from gigl.env.constants import GIGL_RESOURCE_CONFIG_URI_ENV_KEY
 from gigl.env.pipelines_config import get_resource_config
+from gigl.src.common.utils.file_loader import FileLoader
 from tests.test_assets.test_case import TestCase
 
 
@@ -24,8 +28,31 @@ def setUp(self):
             service_account=self._service_account,
             staging_bucket=self._staging_bucket,
         )
+
+        # get_graph_store_info() (run on the launched workers) calls
+        # get_resource_config() to build the readiness URI, so the workers need a
+        # resource config they can read. The test runner's resource config URI may
+        # be a local path that does not exist on the worker image, so we upload the
+        # in-memory resource config to the regional bucket (which the workers can
+        # read from GCS) and pass that URI via GIGL_RESOURCE_CONFIG_URI.
+        self._file_loader = FileLoader()
+        self._remote_resource_config_uri = (
+            self._resource_config.temp_assets_regional_bucket_path
+            / "gigl"
+            / "integration_tests"
+            / "networking"
+            / f"resource_config_{uuid.uuid4()}.yaml"
+        )
+        ProtoUtils().write_proto_to_yaml(
+            proto=self._resource_config.resource_config,
+            uri=self._remote_resource_config_uri,
+        )
         super().setUp()
 
+    def tearDown(self):
+        self._file_loader.delete_files([self._remote_resource_config_uri])
+        super().tearDown()
+
     @parameterized.expand(
         [
             param(
@@ -63,12 +90,22 @@ def test_get_graph_store_info(self, _, storage_nodes, compute_nodes):
                 """
             ),
         ]
+        # launch_graph_store_job propagates the compute pool's environment_variables
+        # to both the compute and storage container specs, so the uploaded resource
+        # config URI is visible to every worker.
+        resource_config_env_vars = [
+            env_var.EnvVar(
+                name=GIGL_RESOURCE_CONFIG_URI_ENV_KEY,
+                value=self._remote_resource_config_uri.uri,
+            )
+        ]
         compute_cluster_config = VertexAiJobConfig(
             job_name=job_name,
             container_uri=DEFAULT_GIGL_RELEASE_SRC_IMAGE_CPU,
             replica_count=compute_nodes,
             command=command,
             machine_type="n2-standard-8",
+            environment_variables=resource_config_env_vars,
         )
         storage_cluster_config = VertexAiJobConfig(
             job_name=job_name,