Merge branch 'master' into py313

.buildkite/_forge.rayci.yml

@@ -25,6 +25,7 @@ steps:
           - "12.1.1-cudnn8"
           - "12.3.2-cudnn9"
           - "12.4.1-cudnn"
+          - "12.5.1-cudnn"
     env:
       PYTHON_VERSION: "{{matrix.python}}"
       CUDA_VERSION: "{{matrix.cuda}}"

.buildkite/build.rayci.yml

@@ -59,7 +59,8 @@ steps:
       - bazel run //ci/ray_ci:build_in_docker -- docker --python-version {{matrix}}
         --platform cu11.7.1-cudnn8 --platform cu11.8.0-cudnn8
         --platform cu12.1.1-cudnn8 --platform cu12.3.2-cudnn9
-        --platform cu12.4.1-cudnn --platform cpu
+        --platform cu12.4.1-cudnn --platform cu12.5.1-cudnn
+        --platform cpu
         --image-type ray --upload
     depends_on:
       - manylinux

.buildkite/linux_aarch64.rayci.yml

@@ -33,6 +33,7 @@ steps:
           - "12.1.1-cudnn8"
           - "12.3.2-cudnn9"
           - "12.4.1-cudnn"
+          - "12.5.1-cudnn"
     instance_type: builder-arm64
     env:
       PYTHON_VERSION: "{{matrix.python}}"
@@ -82,7 +83,8 @@ steps:
       - bazel run //ci/ray_ci:build_in_docker -- docker --python-version {{matrix}}
         --platform cu11.7.1-cudnn8 --platform cu11.8.0-cudnn8
         --platform cu12.1.1-cudnn8 --platform cu12.3.2-cudnn9
-        --platform cu12.4.1-cudnn --platform cpu
+        --platform cu12.4.1-cudnn --platform cu12.5.1-cudnn
+        --platform cpu
         --image-type ray --architecture aarch64 --upload
     depends_on:
       - manylinux-aarch64

.buildkite/llm.rayci.yml

@@ -0,0 +1,17 @@
+group: llm tests
+depends_on:
+  - forge
+  - oss-ci-base_ml
+steps:
+  - name: llmbuild
+    wanda: ci/docker/llm.build.wanda.yaml
+
+  - label: "llm tests"
+    key: "llm-tests"
+    tags:
+      - python
+      - llm
+    instance_type: medium
+    commands:
+      - bazel run //ci/ray_ci:test_in_docker -- //python/ray/llm/... llm
+    depends_on: llmbuild

.buildkite/rllib.rayci.yml

@@ -83,6 +83,7 @@ steps:
     tags:
       - rllib_gpu
       - gpu
+      - skip-on-microcheck
     parallelism: 5
     instance_type: gpu
     commands:
@@ -137,6 +138,7 @@ steps:
     tags:
       - rllib_directly
       - doc
+      - skip-on-microcheck
     instance_type: medium
     commands:
       # doc tests
@@ -159,6 +161,7 @@ steps:
     tags:
       - rllib_gpu
       - gpu
+      - skip-on-microcheck
     parallelism: 5
     instance_type: gpu-large
     commands:

.github/CODEOWNERS

@@ -91,6 +91,9 @@
 /python/ray/train/ @hongpeng-guo @justinvyu @matthewdeng @raulchen @woshiyyya
 /doc/source/train/ @hongpeng-guo @justinvyu @matthewdeng @raulchen @woshiyyya @ray-project/ray-docs
 
+# LLM
+/python/ray/llm/ @ray-project/ray-llm
+
 # Serve (docs)
 /doc/source/serve/ @edoakes @zcin @GeneDer @akshay-anyscale @ray-project/ray-docs
 

.pre-commit-config.yaml

@@ -3,6 +3,8 @@ exclude: |
     python/ray/core/generated/|
     python/ray/serve/generated/|
     python/ray/cloudpickle/|
+    python/ray/dashboard/client/public/|
+    python/ray/tests/test_cli_patterns|
     python/ray/_private/runtime_env/_clonevirtualenv.py|
     doc/external/|
     doc/source/

.rayciversion

@@ -1 +1 @@
-0.8.0
+0.10.0

.vale/styles/config/vocabularies/RLlib/accept.txt

@@ -3,22 +3,26 @@
 [Aa]lgos?
 (APPO|appo)
 [Aa]utoscal(e|ing)
+[Aa]utoregressive
 boolean
 [Cc]allables?
 [Cc]heckpoints?(ing)?
 [Cc]heckpointable
 classmethods?
+CNNs?
 coeff
 config
 (DQN|dqn)
 [Ee]nvs?
 (IMPALA|impala)
 logits?
 log-probs?
+LSTMs?
 hyperparameters?
 MARLModule
 (MARWIL|marwil)
 MLAgents
+MLPs?
 multiagent
 [Pp]erceptrons?
 postprocessing

.vale/styles/config/vocabularies/Train/accept.txt

@@ -5,4 +5,4 @@ LightGBM
 PyTorch
 PyTorch Lightning
 TensorFlow
-XGBoost
+XGBoost

BUILD.bazel

@@ -386,6 +386,7 @@ ray_cc_library(
         "//src/ray/protobuf:common_cc_proto",
         "//src/ray/util",
         "//src/ray/util:compat",
+        "//src/ray/util:counter_map",
         "@msgpack",
     ],
 )
@@ -537,7 +538,10 @@ ray_cc_library(
         ":scheduler",
         ":worker_rpc",
         "//src/ray/protobuf:agent_manager_cc_proto",
+        "//src/ray/util:counter_map",
         "//src/ray/util:thread_checker",
+        "//src/ray/util:throttler",
+        "//src/ray/util:type_traits",
         "@boost//:bimap",
         "@com_github_grpc_grpc//src/proto/grpc/health/v1:health_proto",
         "@com_google_absl//absl/container:btree",
@@ -718,6 +722,7 @@ ray_cc_library(
         "//src/ray/protobuf:common_cc_proto",
         "//src/ray/protobuf:runtime_env_agent_cc_proto",
         "//src/ray/util",
+        "//src/ray/util:throttler",
         "@boost//:asio",
         "@boost//:beast",
         "@boost//:system",
@@ -1972,6 +1977,7 @@ ray_cc_library(
         ":pubsub_lib",
         ":ray_common",
         ":redis_store_client",
+        "//src/ray/util:sequencer",
         "//src/ray/protobuf:usage_cc_proto",
     ],
 )
@@ -2209,6 +2215,7 @@ ray_cc_library(
         ":ray_common",
         ":stats_lib",
         "//src/ray/util",
+        "//src/ray/util:exponential_backoff",
         "@boost//:asio",
     ],
 )
@@ -2224,7 +2231,7 @@ ray_cc_library(
         "src/ray/gcs/store_client/store_client.h",
     ],
     deps = [
-        "redis_client",
+        ":redis_client",
         "@com_google_absl//absl/strings:str_format",
         "@com_google_absl//absl/synchronization",
     ],
@@ -2370,6 +2377,7 @@ ray_cc_library(
         "//src/ray/protobuf:gcs_cc_proto",
         "//src/ray/protobuf:gcs_service_cc_proto",
         "//src/ray/util",
+        "//src/ray/util:exponential_backoff",
         "@boost//:asio",
     ],
 )
@@ -2510,6 +2518,7 @@ pyx_library(
         "//:stats_lib",
         "//src/ray/protobuf:serialization_cc_proto",
         "//src/ray/util",
+        "//src/ray/util:memory",
     ],
 )
 

binder/requirements.txt

@@ -1,4 +1,4 @@
 # Set up requirements needed to build a binder server to launch into.
 -r ../doc/requirements-doc.txt
 -r ../python/requirements.txt
-ray[rllib, serve, tune]
+ray[rllib, serve, tune]

ci/docker/base.gpu.Dockerfile

@@ -18,12 +18,6 @@ ENV RAY_INSTALL_JAVA=0
 ENV BUILDKITE_PULL_REQUEST=${BUILDKITE_PULL_REQUEST}
 ENV BUILDKITE_COMMIT=${BUILDKITE_COMMIT}
 ENV BUILDKITE_PULL_REQUEST_BASE_BRANCH=${BUILDKITE_PULL_REQUEST_BASE_BRANCH}
-# For wheel build
-# https://github.com/docker-library/docker/blob/master/20.10/docker-entrypoint.sh
-ENV DOCKER_TLS_CERTDIR=/certs
-ENV DOCKER_HOST=tcp://docker:2376
-ENV DOCKER_TLS_VERIFY=1
-ENV DOCKER_CERT_PATH=/certs/client
 ENV TRAVIS_COMMIT=${BUILDKITE_COMMIT}
 ENV BUILDKITE_BAZEL_CACHE_URL=${REMOTE_CACHE_URL}
 

ci/docker/base.test.Dockerfile

@@ -14,12 +14,6 @@ ENV PYTHON=$PYTHON
 ENV RAY_USE_RANDOM_PORTS=1
 ENV RAY_DEFAULT_BUILD=1
 ENV RAY_INSTALL_JAVA=0
-# For wheel build
-# https://github.com/docker-library/docker/blob/master/20.10/docker-entrypoint.sh
-ENV DOCKER_TLS_CERTDIR=/certs
-ENV DOCKER_HOST=tcp://docker:2376
-ENV DOCKER_TLS_VERIFY=1
-ENV DOCKER_CERT_PATH=/certs/client
 ENV BUILDKITE_BAZEL_CACHE_URL=${BUILDKITE_BAZEL_CACHE_URL}
 
 RUN <<EOF

ci/docker/core.build.Dockerfile

@@ -3,12 +3,6 @@ FROM $DOCKER_IMAGE_BASE_BUILD
 
 ARG RAYCI_IS_GPU_BUILD=false
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .

ci/docker/data-tfxbsl.build.Dockerfile

@@ -7,12 +7,6 @@ ARG ARROW_VERSION=14.*
 ARG ARROW_MONGO_VERSION=
 ARG RAY_CI_JAVA_BUILD=
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .

ci/docker/data.build.Dockerfile

@@ -7,12 +7,6 @@ ARG ARROW_VERSION=14.*
 ARG ARROW_MONGO_VERSION=
 ARG RAY_CI_JAVA_BUILD=
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .
@@ -35,7 +29,7 @@ sudo apt-get install -y mongodb
 sudo rm -rf /var/lib/mongodb/mongod.lock
 
 if [[ $RAY_CI_JAVA_BUILD == 1 ]]; then
-  # These packages increase the image size quite a bit, so we only install them 
+  # These packages increase the image size quite a bit, so we only install them
   # as needed.
   sudo apt-get install -y -qq maven openjdk-8-jre openjdk-8-jdk
 fi

ci/docker/doc.build.Dockerfile

@@ -1,12 +1,6 @@
 ARG DOCKER_IMAGE_BASE_BUILD=cr.ray.io/rayproject/oss-ci-base_build
 FROM $DOCKER_IMAGE_BASE_BUILD
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .

ci/docker/docgpu.build.Dockerfile

@@ -3,12 +3,6 @@
 ARG DOCKER_IMAGE_BASE_BUILD=cr.ray.io/rayproject/oss-ci-base_gpu
 FROM $DOCKER_IMAGE_BASE_BUILD
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .

ci/docker/doctest.build.Dockerfile

@@ -1,12 +1,6 @@
 ARG DOCKER_IMAGE_BASE_BUILD=cr.ray.io/rayproject/oss-ci-base_build
 FROM $DOCKER_IMAGE_BASE_BUILD
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .

ci/docker/forge.Dockerfile

@@ -1,6 +1,6 @@
 # syntax=docker/dockerfile:1.3-labs
 
-FROM ubuntu:20.04
+FROM ubuntu:22.04
 
 ARG BUILDKITE_BAZEL_CACHE_URL
 
@@ -40,7 +40,7 @@ apt-get install -y \
 
 # As a convention, we pin all python packages to a specific version. This
 # is to to make sure we can control version upgrades through code changes.
-python -m pip install pip==23.2.1 cffi==1.16.0
+python -m pip install pip==25.0 cffi==1.16.0
 
 # Needs to be synchronized to the host group id as we map /var/run/docker.sock
 # into the container.

ci/docker/forge.aarch64.wanda.yaml

@@ -1,6 +1,6 @@
 name: "forge-aarch64"
 froms:
-  - ubuntu:20.04
+  - ubuntu:22.04
 build_args:
   - BUILDKITE_BAZEL_CACHE_URL
 dockerfile: ci/docker/forge.Dockerfile

ci/docker/forge.wanda.yaml

@@ -1,6 +1,6 @@
 name: "forge"
 froms:
-  - ubuntu:20.04
+  - ubuntu:22.04
 build_args:
   - BUILDKITE_BAZEL_CACHE_URL
 dockerfile: ci/docker/forge.Dockerfile

ci/docker/llm.build.Dockerfile

@@ -0,0 +1,20 @@
+# syntax=docker/dockerfile:1.3-labs
+
+ARG DOCKER_IMAGE_BASE_BUILD=cr.ray.io/rayproject/oss-ci-base_ml
+FROM $DOCKER_IMAGE_BASE_BUILD
+
+ARG RAY_CI_JAVA_BUILD=
+
+SHELL ["/bin/bash", "-ice"]
+
+COPY . .
+
+RUN <<EOF
+#!/bin/bash
+
+set -ex
+
+# TODO(comaniac): add other test dependencies here.
+pip install -c python/requirements_compiled.txt pytest aiohttp pillow
+
+EOF

ci/docker/llm.build.wanda.yaml

@@ -0,0 +1,9 @@
+name: "llmbuild"
+froms: ["cr.ray.io/rayproject/oss-ci-base_ml"]
+dockerfile: ci/docker/llm.build.Dockerfile
+srcs:
+  - ci/env/install-dependencies.sh
+  - python/requirements.txt
+  - python/requirements_compiled.txt
+tags:
+  - cr.ray.io/rayproject/llmbuild

ci/docker/min.build.Dockerfile

@@ -6,12 +6,6 @@ FROM $DOCKER_IMAGE_BASE_BUILD
 ARG PYTHON_VERSION
 ARG EXTRA_DEPENDENCY
 
-# Unset dind settings; we are using the host's docker daemon.
-ENV DOCKER_TLS_CERTDIR=
-ENV DOCKER_HOST=
-ENV DOCKER_TLS_VERIFY=
-ENV DOCKER_CERT_PATH=
-
 SHELL ["/bin/bash", "-ice"]
 
 COPY . .