[v1.x] Add new CI pipeline for building and testing with cuda 11.0.

ci/docker/Dockerfile.build.ubuntu_build_cuda110

@@ -0,0 +1,46 @@
+# -*- mode: dockerfile -*-
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+#
+# Dockerfile to build MXNet on Ubuntu 16.04 for GPU but on
+# a CPU-only instance. This restriction is caused by the CPP-
+# package generation, requiring the actual CUDA library to be
+# present
+
+FROM nvidia/cuda:11.0-cudnn8-devel-ubuntu16.04
+
+WORKDIR /work/deps
+
+COPY install/ubuntu_core.sh /work/
+RUN /work/ubuntu_core.sh
+COPY install/deb_ubuntu_ccache.sh /work/
+RUN /work/deb_ubuntu_ccache.sh
+COPY install/ubuntu_python.sh /work/
+COPY install/requirements /work/
+RUN /work/ubuntu_python.sh
+
+# Keep this at the end since this command is not cachable
+ARG USER_ID=0
+ARG GROUP_ID=0
+COPY install/ubuntu_adduser.sh /work/
+RUN /work/ubuntu_adduser.sh
+
+COPY runtime_functions.sh /work/
+
+WORKDIR /work/mxnet
+ENV LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/compat
+

ci/docker/Dockerfile.build.ubuntu_gpu_cu110

@@ -0,0 +1,44 @@
+# -*- mode: dockerfile -*-
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+#
+# Dockerfile to run MXNet on Ubuntu 16.04 for GPU
+
+FROM nvidia/cuda:11.0-cudnn8-devel-ubuntu16.04
+
+WORKDIR /work/deps
+
+COPY install/ubuntu_core.sh /work/
+RUN /work/ubuntu_core.sh
+
+COPY install/deb_ubuntu_ccache.sh /work/
+RUN /work/deb_ubuntu_ccache.sh
+
+COPY install/ubuntu_python.sh /work/
+COPY install/requirements /work/
+RUN /work/ubuntu_python.sh
+
+# Always last
+ARG USER_ID=0
+ARG GROUP_ID=0
+COPY install/ubuntu_adduser.sh /work/
+RUN /work/ubuntu_adduser.sh
+
+COPY runtime_functions.sh /work/
+
+WORKDIR /work/mxnet
+ENV LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/compat

ci/docker/install/ubuntu_cudnn.sh

@@ -32,31 +32,41 @@ fi
 apt-get update || true
 
 case ${CUDA_VERSION} in
+	11\.0*)
+		export libcudnn_package="libcudnn8"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda11.0"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda11.0"
+		;;
 	10\.2*)
-		export libcudnn7_version="${CUDNN_VERSION}-1+cuda10.2"
-		export libcudnn7_dev_version="${CUDNN_VERSION}-1+cuda10.2"
+		export libcudnn_package="libcudnn7"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda10.2"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda10.2"
 		;;
 	10\.1*)
-		export libcudnn7_version="${CUDNN_VERSION}-1+cuda10.1"
-		export libcudnn7_dev_version="${CUDNN_VERSION}-1+cuda10.1"
+		export libcudnn_package="libcudnn7"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda10.1"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda10.1"
 		;;
 	10\.0*)
-		export libcudnn7_version="${CUDNN_VERSION}-1+cuda10.0"
-		export libcudnn7_dev_version="${CUDNN_VERSION}-1+cuda10.0"
+		export libcudnn_package="libcudnn7"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda10.0"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda10.0"
 		;;
 	9\.0*)
-		export libcudnn7_version="${CUDNN_VERSION}-1+cuda9.0"
-		export libcudnn7_dev_version="${CUDNN_VERSION}-1+cuda9.0"
+		export libcudnn_package="libcudnn7"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda9.0"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda9.0"
 		;;
 	9\.2*)
-		export libcudnn7_version="${CUDNN_VERSION}-1+cuda9.2"
-		export libcudnn7_dev_version="${CUDNN_VERSION}-1+cuda9.2"
+		export libcudnn_package="libcudnn7"
+		export libcudnn_version="${CUDNN_VERSION}-1+cuda9.2"
+		export libcudnn_dev_version="${CUDNN_VERSION}-1+cuda9.2"
 		;;
 	*)
 		echo "Unsupported CUDA version ${CUDA_VERSION}"
 		exit 1
 		;;
 esac
 
-apt-get install -y --allow-downgrades libcudnn7=${libcudnn7_version} libcudnn7-dev=${libcudnn7_dev_version}
+apt-get install -y --allow-downgrades ${libcudnn_package}=${libcudnn_version} ${libcudnn_package}-dev=${libcudnn_dev_version}
 

ci/docker/runtime_functions.sh

@@ -806,6 +806,24 @@ build_ubuntu_gpu_cuda101_cudnn7() {
     make cython PYTHON=python3
 }
 
+build_ubuntu_gpu_cuda110_cudnn8() {
+    set -ex
+    build_ccache_wrappers
+    make \
+        USE_BLAS=openblas                         \
+        USE_MKLDNN=0                              \
+        USE_CUDA=1                                \
+        USE_CUDA_PATH=/usr/local/cuda             \
+        USE_CUDNN=1                               \
+        USE_TVM_OP=0                              \
+        USE_CPP_PACKAGE=1                         \
+        USE_DIST_KVSTORE=1                        \
+        CUDA_ARCH="$CI_CUDA_COMPUTE_CAPABILITIES" \
+        USE_SIGNAL_HANDLER=1                      \
+        -j$(nproc)
+    make cython PYTHON=python3
+}
+
 build_ubuntu_gpu_cuda101_cudnn7_mkldnn_cpp_test() {
     set -ex
     build_ccache_wrappers
@@ -1083,6 +1101,18 @@ unittest_ubuntu_python3_quantization_gpu() {
     nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_quantization_gpu.xml --verbose tests/python/quantization_gpu
 }
 
+unittest_ubuntu_python3_quantization_gpu_cu110() {
+    set -ex
+    export PYTHONPATH=./python/
+    export MXNET_MKLDNN_DEBUG=0 # Ignored if not present
+    export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
+    export MXNET_SUBGRAPH_VERBOSE=0
+    export CUDNN_VERSION=${CUDNN_VERSION:-8.0.33}
+    export MXNET_ENABLE_CYTHON=0
+    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_quantization_gpu.xml --verbose tests/python/quantization_gpu
+}
+
 unittest_centos7_cpu_scala() {
     set -ex
     cd /work/mxnet

ci/jenkins/Jenkins_steps.groovy

@@ -239,6 +239,20 @@ def compile_unix_full_gpu() {
     }]
 }
 
+def compile_unix_full_gpu_cu110() {
+    return ['GPU: CUDA11.0+cuDNN8': {
+      node(NODE_LINUX_CPU) {
+        ws('workspace/build-gpu') {
+          timeout(time: max_time, unit: 'MINUTES') {
+            utils.init_git()
+            utils.docker_run('ubuntu_build_cuda110', 'build_ubuntu_gpu_cuda110_cudnn8', false)
+            utils.pack_lib('gpu_cu110', mx_lib_cpp_examples)
+          }
+        }
+      }
+    }]
+}
+
 def compile_unix_full_gpu_mkldnn_cpp_test() {
     return ['GPU: CUDA10.1+cuDNN7+MKLDNN+CPPTEST': {
       node(NODE_LINUX_CPU) {
@@ -737,6 +751,22 @@ def test_unix_python3_gpu() {
     }]
 }
 
+def test_unix_python3_gpu_cu110() {
+    return ['Python3+CUDA11.0: GPU': {
+      node(NODE_LINUX_GPU_G4) {
+        ws('workspace/ut-python3-gpu') {
+          try {
+            utils.unpack_and_init('gpu_cu110', mx_lib_cython)
+            python3_gpu_ut_cython('ubuntu_gpu_cu110')
+            utils.publish_test_coverage()
+          } finally {
+            utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_gpu.xml')
+          }
+        }
+      }
+    }]
+}
+
 def test_unix_python3_quantize_gpu() {
     return ['Python3: Quantize GPU': {
       node(NODE_LINUX_GPU_P3) {
@@ -755,6 +785,24 @@ def test_unix_python3_quantize_gpu() {
     }]
 }
 
+def test_unix_python3_quantize_gpu_cu110() {
+    return ['Python3+CUDA11.0: Quantize GPU': {
+      node(NODE_LINUX_GPU_P3) {
+        ws('workspace/ut-python3-quantize-gpu') {
+          timeout(time: max_time, unit: 'MINUTES') {
+            try {
+              utils.unpack_and_init('gpu_cu110', mx_lib)
+              utils.docker_run('ubuntu_gpu_cu110', 'unittest_ubuntu_python3_quantization_gpu', true)
+              utils.publish_test_coverage()
+            } finally {
+              utils.collect_test_results_unix('nosetests_quantization_gpu.xml', 'nosetests_python3_quantize_gpu.xml')
+            }
+          }
+        }
+      }
+    }]
+}
+
 def test_unix_python3_debug_cpu() {
     return ['Python3: CPU debug': {
       node(NODE_LINUX_CPU) {

ci/jenkins/Jenkinsfile_full

@@ -31,6 +31,7 @@ def buildJobs = [
     'miscellaneous',
     'unix-cpu',
     'unix-gpu',
+    'unix-gpu-cu110',
     'website',
     'windows-cpu',
     'windows-gpu'

ci/jenkins/Jenkinsfile_unix_gpu_cu110

@@ -0,0 +1,52 @@
+// -*- mode: groovy -*-
+
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+//
+// Jenkins pipeline
+// See documents at https://jenkins.io/doc/book/pipeline/jenkinsfile/
+
+// timeout in minutes
+max_time = 180
+
+node('utility') {
+  // Loading the utilities requires a node context unfortunately
+  checkout scm
+  utils = load('ci/Jenkinsfile_utils.groovy')
+  custom_steps = load('ci/jenkins/Jenkins_steps.groovy')
+}
+utils.assign_node_labels(utility: 'utility', linux_cpu: 'mxnetlinux-cpu', linux_gpu: 'mxnetlinux-gpu', linux_gpu_p3: 'mxnetlinux-gpu-p3', linux_gpu_g4: 'mxnetlinux-gpu-g4')
+
+utils.main_wrapper(
+core_logic: {
+  utils.parallel_stage('Build', [
+    custom_steps.compile_unix_full_gpu_cu110()
+  ])
+
+  utils.parallel_stage('Tests', [
+    custom_steps.test_unix_python3_gpu_cu110(),
+    custom_steps.test_unix_python3_quantize_gpu_cu110()
+  ]) 
+}
+,
+failure_handler: {
+  // Only send email if master or release branches failed
+  if (currentBuild.result == "FAILURE" && (env.BRANCH_NAME == "master" || env.BRANCH_NAME.startsWith("v"))) {
+    emailext body: 'Build for MXNet branch ${BRANCH_NAME} has broken. Please view the build at ${BUILD_URL}', replyTo: '${EMAIL}', subject: '[BUILD FAILED] Branch ${BRANCH_NAME} build ${BUILD_NUMBER}', to: '${EMAIL}'
+  }
+}
+)