diff --git a/3rdparty/mshadow/mshadow/tensor.h b/3rdparty/mshadow/mshadow/tensor.h
index f8f63388155c..8b24d1fa6136 100644
--- a/3rdparty/mshadow/mshadow/tensor.h
+++ b/3rdparty/mshadow/mshadow/tensor.h
@@ -108,7 +108,10 @@ struct Shape {
    * \return the corresponding dimension size
    */
   MSHADOW_XINLINE const index_t &operator[](int idx) const {
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Warray-bounds"
     return shape_[idx];
+#pragma GCC diagnostic pop
   }
   /*!
    * \return whether two shape equals
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 7d34d5f1935e..dd1206adeab9 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -77,7 +77,6 @@ option(USE_JEMALLOC "Build with Jemalloc support" OFF)
 option(USE_LIBJPEG_TURBO "Use libjpeg-turbo" OFF)
 option(USE_DIST_KVSTORE "Build with DIST_KVSTORE support" OFF)
 option(USE_PLUGINS_WARPCTC "Use WARPCTC Plugins" OFF)
-option(USE_CPP_PACKAGE "Build C++ Package" OFF)
 option(USE_MXNET_LIB_NAMING "Use MXNet library naming conventions." ON)
 option(USE_GPROF "Compile with gprof (profiling) flag" OFF)
 option(USE_VTUNE "Enable use of Intel Amplifier XE (VTune)" OFF) # one could set VTUNE_ROOT for search path
diff --git a/ci/docker/Dockerfile.build.ubuntu b/ci/docker/Dockerfile.build.ubuntu
index e68e2f94bc02..73494a6784a8 100644
--- a/ci/docker/Dockerfile.build.ubuntu
+++ b/ci/docker/Dockerfile.build.ubuntu
@@ -21,8 +21,8 @@
 # See docker-compose.yml for supported BASE_IMAGE ARGs and targets.
 
 ####################################################################################################
-# The Dockerfile uses a dynamic BASE_IMAGE (for example ubuntu:18.04
-# nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04 etc).
+# The Dockerfile uses a dynamic BASE_IMAGE (for example ubuntu:20.04
+# nvidia/cuda:11.1-cudnn8-devel-ubuntu20.04 etc).
 # On top of BASE_IMAGE we install all dependencies shared by all MXNet build
 # environments into a "base" target. At the end of this file, we can specialize
 # "base" for specific usecases. The target built by docker can be selected via
@@ -36,9 +36,7 @@ WORKDIR /work/deps
 RUN export DEBIAN_FRONTEND=noninteractive && \
     apt-get update && \
     apt-get install -y wget software-properties-common && \
-    wget -qO - http://apt.llvm.org/llvm-snapshot.gpg.key | apt-key add - && \
     wget -qO - wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB | apt-key add - && \
-    apt-add-repository "deb http://apt.llvm.org/xenial/ llvm-toolchain-xenial-10 main" &&  \
     apt-add-repository "deb https://apt.repos.intel.com/mkl all main" &&  \
     apt-get update && \
     apt-get install -y \
@@ -47,6 +45,7 @@ RUN export DEBIAN_FRONTEND=noninteractive && \
         unzip \
         pandoc \
         ## Development tools
+        cmake \
         build-essential \
         ninja-build \
         git \
@@ -123,43 +122,33 @@ COPY runtime_functions.sh /work/
 # The target built by docker can be selected via "--target" option or docker-compose.yml
 ####################################################################################################
 FROM base as gpu
-# Install Thrust 1.9.8 to be shipped with Cuda 11.
-# Fixes https://github.com/thrust/thrust/issues/1072 for Clang 10
-# This file can be deleted when using Cuda 11 on CI
-RUN cd /usr/local && \
-    git clone https://github.com/thrust/thrust.git && \
-    cd thrust && \
-    git checkout 1.9.8
 
-# Install TensorRT
+# Install TensorRT and CuDNN
+# Use bash as it has better support for string comparisons in if clauses
+SHELL ["/bin/bash", "-c"]
 # We need to redeclare ARG due to
 # https://docs.docker.com/engine/reference/builder/#understand-how-arg-and-from-interact
 ARG BASE_IMAGE
-# due to issue https://gitlab.com/nvidia/container-images/cuda/-/issues/92, we
-#  get the cuda version from installed deb package if CUDA_VERSION is not set
-RUN if [ -z "$CUDA_VERSION" ]; then export CUDA_VERSION=$( \
-    dpkg --list cuda-libraries*| grep "cuda-libraries-dev-" | awk '{print $3}' | sed 's/-1$//'); \
-    fi && export SHORT_CUDA_VERSION=${CUDA_VERSION%.*} && \
+RUN export SHORT_CUDA_VERSION=${CUDA_VERSION%.*} && \
+    export OS_RELEASE="$(cat /etc/os-release)" && \
     apt-get update && \
-    if [ ${SHORT_CUDA_VERSION} = 10.0 ]; then \
-        TRT_VERSION="7.0.0-1+cuda10.0"; \
-        TRT_MAJOR_VERSION=7; \
-    elif [ ${SHORT_CUDA_VERSION} = 10.1 ]; then \
-        TRT_VERSION="6.0.1-1+cuda10.1"; \
-        TRT_MAJOR_VERSION=6; \
-    elif [ ${SHORT_CUDA_VERSION} = 10.2 ]; then \
-        TRT_VERSION="7.0.0-1+cuda10.2"; \
-        TRT_MAJOR_VERSION=7; \
-    elif [ ${SHORT_CUDA_VERSION} = 11.0 ]; then \
-        TRT_VERSION="7.2.0-1+cuda11.0"; \
-        TRT_MAJOR_VERSION=7; \
-    else \
-        echo "ERROR: Cuda ${SHORT_CUDA_VERSION} not yet supported in Dockerfile.build.ubuntu"; \
-        exit 1; \
+    if [[ ${OS_RELEASE} == *"Bionic"* ]]; then \
+        if [ ${SHORT_CUDA_VERSION} = 11.0 ]; then \
+            TRT_VERSION="7.2.0-1+cuda11.0"; \
+            TRT_MAJOR_VERSION=7; \
+        elif [ ${SHORT_CUDA_VERSION} = 11.1 ]; then \
+            TRT_VERSION="7.2.1-1+cuda11.1"; \
+            TRT_MAJOR_VERSION=7; \
+        else \
+            echo "ERROR: Cuda ${SHORT_CUDA_VERSION} not yet supported in Dockerfile.build.ubuntu"; \
+            exit 1; \
+        fi; \
+        apt-get install -y libnvinfer${TRT_MAJOR_VERSION}=${TRT_VERSION} \
+                           libnvinfer-dev=${TRT_VERSION} \
+                           libnvinfer-plugin${TRT_MAJOR_VERSION}=${TRT_VERSION} \
+                           libnvinfer-plugin-dev=${TRT_VERSION}; \
     fi && \
-    apt-get install -y libnvinfer${TRT_MAJOR_VERSION}=${TRT_VERSION} \
-                       libnvinfer-dev=${TRT_VERSION} \
-                       libnvinfer-plugin${TRT_MAJOR_VERSION}=${TRT_VERSION} \
-                       libnvinfer-plugin-dev=${TRT_VERSION} && \
+    apt-get install -y libcudnn8-dev && \
     rm -rf /var/lib/apt/lists/*
 
+ENV CUDNN_VERSION=8.0.5
diff --git a/ci/docker/docker-compose.yml b/ci/docker/docker-compose.yml
index c7a11c499fdf..6ebc6a0e025e 100644
--- a/ci/docker/docker-compose.yml
+++ b/ci/docker/docker-compose.yml
@@ -85,49 +85,29 @@ services:
       dockerfile: Dockerfile.build.ubuntu
       target: base
       args:
-        BASE_IMAGE: ubuntu:18.04
+        BASE_IMAGE: ubuntu:20.04
       cache_from:
         - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_cpu:latest
-  ubuntu_gpu_cu101:
-    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu101:latest
+  ubuntu_tensorrt_cu111:
+    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_tensorrt_cu111:latest
     build:
       context: .
       dockerfile: Dockerfile.build.ubuntu
       target: gpu
       args:
-        BASE_IMAGE: nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04
+        BASE_IMAGE: nvidia/cuda:11.1-devel-ubuntu18.04
       cache_from:
-        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu101:latest
-  ubuntu_gpu_cu102:
-    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu102:latest
+        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_tensorrt_cu111:latest
+  ubuntu_gpu_cu111:
+    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu111:latest
     build:
       context: .
       dockerfile: Dockerfile.build.ubuntu
       target: gpu
       args:
-        BASE_IMAGE: nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04
+        BASE_IMAGE: nvidia/cuda:11.1-devel-ubuntu20.04
       cache_from:
-        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu102:latest
-  ubuntu_gpu_cu110:
-    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu110:latest
-    build:
-      context: .
-      dockerfile: Dockerfile.build.ubuntu
-      target: gpu
-      args:
-        BASE_IMAGE: nvidia/cuda:11.0-cudnn8-devel-ubuntu18.04
-      cache_from:
-        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu110:latest
-  ubuntu_build_cuda:
-    image: ${DOCKER_CACHE_REGISTRY}/build.ubuntu_build_cuda:latest
-    build:
-      context: .
-      dockerfile: Dockerfile.build.ubuntu
-      target: gpu
-      args:
-        BASE_IMAGE: nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04
-      cache_from:
-        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_build_cuda:latest
+        - ${DOCKER_CACHE_REGISTRY}/build.ubuntu_gpu_cu111:latest
   ###################################################################################################
   # Dockerfile.build.android based images used for testing cross-compilation for plain ARM
   ###################################################################################################
diff --git a/ci/docker/runtime_functions.sh b/ci/docker/runtime_functions.sh
index 88e77a77a73e..c8ec09a33943 100755
--- a/ci/docker/runtime_functions.sh
+++ b/ci/docker/runtime_functions.sh
@@ -314,11 +314,10 @@ build_ubuntu_cpu() {
 build_ubuntu_cpu_openblas() {
     set -ex
     cd /work/build
-    CXXFLAGS="-Wno-error=strict-overflow" CC=gcc-7 CXX=g++-7 cmake \
+    CXXFLAGS="-Wno-error=strict-overflow" cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DENABLE_TESTCOVERAGE=ON \
         -DUSE_TVM_OP=ON \
-        -DUSE_CPP_PACKAGE=ON \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_MKLDNN=OFF \
         -DUSE_CUDA=OFF \
@@ -332,7 +331,7 @@ build_ubuntu_cpu_openblas() {
 build_ubuntu_cpu_mkl() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DENABLE_TESTCOVERAGE=OFF \
         -DUSE_MKLDNN=OFF \
@@ -348,7 +347,7 @@ build_ubuntu_cpu_mkl() {
 build_ubuntu_cpu_cmake_debug() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE=Debug \
         -DENABLE_TESTCOVERAGE=ON \
         -DUSE_CUDA=OFF \
@@ -365,7 +364,7 @@ build_ubuntu_cpu_cmake_debug() {
 build_ubuntu_cpu_cmake_no_tvm_op() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DUSE_CUDA=OFF \
         -DUSE_TVM_OP=OFF \
         -DUSE_MKL_IF_AVAILABLE=OFF \
@@ -384,8 +383,6 @@ build_ubuntu_cpu_cmake_asan() {
     set -ex
 
     cd /work/build
-    export CXX=g++-8
-    export CC=gcc-8
     cmake \
         -DUSE_CUDA=OFF \
         -DUSE_MKL_IF_AVAILABLE=OFF \
@@ -396,8 +393,6 @@ build_ubuntu_cpu_cmake_asan() {
         -DUSE_GPERFTOOLS=OFF \
         -DUSE_JEMALLOC=OFF \
         -DUSE_ASAN=ON \
-        -DUSE_CPP_PACKAGE=ON \
-        -DMXNET_USE_CPU=ON \
         /work/mxnet
     make -j $(nproc) mxnet
 }
@@ -405,11 +400,9 @@ build_ubuntu_cpu_cmake_asan() {
 build_ubuntu_cpu_gcc8_werror() {
     set -ex
     cd /work/build
-    CXX=g++-8 CC=gcc-8 cmake \
+    CC=gcc-8 CXX=g++-8 cmake \
         -DUSE_CUDA=OFF \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
-        -DUSE_CPP_PACKAGE=ON \
-        -DMXNET_USE_CPU=ON \
         -GNinja /work/mxnet
     ninja
 }
@@ -420,8 +413,6 @@ build_ubuntu_cpu_clang10_werror() {
     CXX=clang++-10 CC=clang-10 cmake \
        -DUSE_CUDA=OFF \
        -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
-       -DUSE_CPP_PACKAGE=ON \
-       -DMXNET_USE_CPU=ON \
        -GNinja /work/mxnet
     ninja
 }
@@ -438,9 +429,9 @@ build_ubuntu_gpu_clang10_werror() {
 
     CXX=clang++-10 CC=clang-10 cmake \
        -DUSE_CUDA=ON \
+       -DUSE_NVML=OFF \
        -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
        -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
-       -DUSE_CPP_PACKAGE=OFF \
        -GNinja /work/mxnet
     ninja
 }
@@ -454,7 +445,6 @@ build_ubuntu_cpu_clang6() {
         -DUSE_CUDA=OFF \
         -DUSE_OPENMP=OFF \
         -DUSE_DIST_KVSTORE=ON \
-        -DUSE_CPP_PACKAGE=ON \
         -G Ninja /work/mxnet
     ninja
 }
@@ -468,7 +458,6 @@ build_ubuntu_cpu_clang100() {
        -DUSE_CUDA=OFF \
        -DUSE_OPENMP=ON \
        -DUSE_DIST_KVSTORE=ON \
-       -DUSE_CPP_PACKAGE=ON \
        -G Ninja /work/mxnet
     ninja
 }
@@ -484,7 +473,6 @@ build_ubuntu_cpu_clang_tidy() {
        -DUSE_OPENMP=OFF \
        -DCMAKE_BUILD_TYPE=Debug \
        -DUSE_DIST_KVSTORE=ON \
-       -DUSE_CPP_PACKAGE=ON \
        -DCMAKE_CXX_CLANG_TIDY=clang-tidy-10 \
        -G Ninja /work/mxnet
     ninja
@@ -497,7 +485,6 @@ build_ubuntu_cpu_clang6_mkldnn() {
        -DUSE_MKL_IF_AVAILABLE=OFF \
        -DUSE_MKLDNN=ON \
        -DUSE_CUDA=OFF \
-       -DUSE_CPP_PACKAGE=ON \
        -DUSE_OPENMP=OFF \
        -G Ninja /work/mxnet
     ninja
@@ -510,7 +497,6 @@ build_ubuntu_cpu_clang100_mkldnn() {
        -DUSE_MKL_IF_AVAILABLE=OFF \
        -DUSE_MKLDNN=ON \
        -DUSE_CUDA=OFF \
-       -DUSE_CPP_PACKAGE=ON \
        -G Ninja /work/mxnet
     ninja
 }
@@ -518,14 +504,13 @@ build_ubuntu_cpu_clang100_mkldnn() {
 build_ubuntu_cpu_mkldnn() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DENABLE_TESTCOVERAGE=ON \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_TVM_OP=ON \
         -DUSE_MKLDNN=ON \
         -DUSE_CUDA=OFF \
-        -DUSE_CPP_PACKAGE=ON \
         -DBUILD_EXTENSION_PATH=/work/mxnet/example/extensions/lib_external_ops \
         -G Ninja /work/mxnet
     ninja
@@ -534,7 +519,7 @@ build_ubuntu_cpu_mkldnn() {
 build_ubuntu_cpu_mkldnn_mkl() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DENABLE_TESTCOVERAGE=OFF \
         -DUSE_MKLDNN=ON \
@@ -547,16 +532,10 @@ build_ubuntu_cpu_mkldnn_mkl() {
     ninja
 }
 
-build_ubuntu_gpu() {
-    build_ubuntu_gpu_cuda101_cudnn7
-}
-
 build_ubuntu_gpu_tensorrt() {
 
     set -ex
 
-    export CC=gcc-7
-    export CXX=g++-7
     export ONNX_NAMESPACE=onnx
 
     # Build ONNX
@@ -597,6 +576,7 @@ build_ubuntu_gpu_tensorrt() {
           -DUSE_TENSORRT=1                        \
           -DUSE_OPENMP=0                          \
           -DUSE_MKLDNN=0                          \
+          -DUSE_NVML=OFF                          \
           -DUSE_MKL_IF_AVAILABLE=OFF              \
           -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
           -G Ninja                                \
@@ -608,12 +588,12 @@ build_ubuntu_gpu_tensorrt() {
 build_ubuntu_gpu_mkldnn() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_CUDA=ON \
+        -DUSE_NVML=OFF \
         -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
-        -DUSE_CPP_PACKAGE=ON \
         -DBUILD_EXTENSION_PATH=/work/mxnet/example/extensions/lib_external_ops \
         -G Ninja /work/mxnet
     ninja
@@ -622,29 +602,29 @@ build_ubuntu_gpu_mkldnn() {
 build_ubuntu_gpu_mkldnn_nocudnn() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_CUDA=ON \
+        -DUSE_NVML=OFF \
         -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
         -DUSE_CUDNN=OFF \
-        -DUSE_CPP_PACKAGE=ON \
         -DBUILD_EXTENSION_PATH=/work/mxnet/example/extensions/lib_external_ops \
         -G Ninja /work/mxnet
     ninja
 }
 
-build_ubuntu_gpu_cuda101_cudnn7() {
+build_ubuntu_gpu() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE="RelWithDebInfo" \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_CUDA=ON \
+        -DUSE_NVML=OFF \
         -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
         -DUSE_CUDNN=ON \
         -DUSE_MKLDNN=OFF \
-        -DUSE_CPP_PACKAGE=ON \
         -DUSE_DIST_KVSTORE=ON \
         -DBUILD_CYTHON_MODULES=ON \
         -DBUILD_EXTENSION_PATH=/work/mxnet/example/extensions/lib_external_ops \
@@ -652,47 +632,27 @@ build_ubuntu_gpu_cuda101_cudnn7() {
     ninja
 }
 
-build_ubuntu_gpu_cuda101_cudnn7_debug() {
+build_ubuntu_gpu_debug() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DCMAKE_BUILD_TYPE=Debug \
         -DUSE_MKL_IF_AVAILABLE=OFF \
         -DUSE_CUDA=ON \
+        -DUSE_NVML=OFF \
         -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
         -DUSE_CUDNN=ON \
         -DUSE_MKLDNN=OFF \
-        -DUSE_CPP_PACKAGE=ON \
         -DUSE_DIST_KVSTORE=ON \
         -DBUILD_CYTHON_MODULES=ON \
         -G Ninja /work/mxnet
     ninja
 }
 
-build_ubuntu_gpu_cmake() {
-    set -ex
-    cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
-        -DUSE_SIGNAL_HANDLER=ON                 \
-        -DUSE_CUDA=ON                           \
-        -DUSE_CUDNN=ON                          \
-        -DUSE_MKL_IF_AVAILABLE=OFF              \
-        -DUSE_MKLML_MKL=OFF                     \
-        -DUSE_MKLDNN=OFF                        \
-        -DUSE_DIST_KVSTORE=ON                   \
-        -DCMAKE_BUILD_TYPE=Release              \
-        -DMXNET_CUDA_ARCH="$CI_CMAKE_CUDA_ARCH" \
-        -DBUILD_CYTHON_MODULES=1                \
-        -G Ninja                                \
-        /work/mxnet
-
-    ninja
-}
-
 build_ubuntu_cpu_large_tensor() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DUSE_SIGNAL_HANDLER=ON                 \
         -DUSE_CUDA=OFF                          \
         -DUSE_CUDNN=OFF                         \
@@ -706,12 +666,12 @@ build_ubuntu_cpu_large_tensor() {
 build_ubuntu_gpu_large_tensor() {
     set -ex
     cd /work/build
-    CC=gcc-7 CXX=g++-7 cmake \
+    cmake \
         -DUSE_SIGNAL_HANDLER=ON                 \
         -DUSE_CUDA=ON                           \
         -DUSE_CUDNN=ON                          \
+        -DUSE_NVML=OFF \
         -DUSE_MKL_IF_AVAILABLE=OFF              \
-        -DUSE_MKLML_MKL=OFF                     \
         -DUSE_MKLDNN=ON                         \
         -DUSE_DIST_KVSTORE=ON                   \
         -DCMAKE_BUILD_TYPE=Release              \
diff --git a/ci/jenkins/Jenkins_steps.groovy b/ci/jenkins/Jenkins_steps.groovy
index fca8e44b25ff..1413756dc2c8 100644
--- a/ci/jenkins/Jenkins_steps.groovy
+++ b/ci/jenkins/Jenkins_steps.groovy
@@ -139,7 +139,7 @@ def compile_unix_int64_gpu(lib_name) {
         ws('workspace/build-gpu-int64') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_gpu_cu101', 'build_ubuntu_gpu_large_tensor', false)
+            utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu_large_tensor', false)
             utils.pack_lib(lib_name, mx_cmake_lib)
           }
         }
@@ -195,7 +195,7 @@ def compile_unix_mkldnn_gpu(lib_name) {
         ws('workspace/build-mkldnn-gpu') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_build_cuda', 'build_ubuntu_gpu_mkldnn', false)
+            utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu_mkldnn', false)
             utils.pack_lib(lib_name, mx_mkldnn_lib)
           }
         }
@@ -209,7 +209,7 @@ def compile_unix_mkldnn_nocudnn_gpu(lib_name) {
          ws('workspace/build-mkldnn-gpu-nocudnn') {
            timeout(time: max_time, unit: 'MINUTES') {
              utils.init_git()
-             utils.docker_run('ubuntu_build_cuda', 'build_ubuntu_gpu_mkldnn_nocudnn', false)
+             utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu_mkldnn_nocudnn', false)
              utils.pack_lib(lib_name, mx_mkldnn_lib)
            }
          }
@@ -218,12 +218,12 @@ def compile_unix_mkldnn_nocudnn_gpu(lib_name) {
 }
 
 def compile_unix_full_gpu(lib_name) {
-    return ['GPU: CUDA10.1+cuDNN7': {
+    return ['GPU: CUDA+cuDNN': {
       node(NODE_LINUX_CPU) {
         ws('workspace/build-gpu') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_build_cuda', 'build_ubuntu_gpu_cuda101_cudnn7', false)
+            utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu', false)
             utils.pack_lib(lib_name, mx_lib_cpp_examples)
           }
         }
@@ -232,12 +232,12 @@ def compile_unix_full_gpu(lib_name) {
 }
 
 def compile_unix_full_gpu_debug(lib_name) {
-    return ['GPU: CUDA10.1+cuDNN7, debug': {
+    return ['GPU: CUDA+cuDNN, debug': {
       node(NODE_LINUX_CPU) {
         ws('workspace/build-gpu') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_build_cuda', 'build_ubuntu_gpu_cuda101_cudnn7_debug', false)
+            utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu_debug', false)
             utils.pack_lib(lib_name, mx_lib_cpp_examples)
           }
         }
@@ -245,27 +245,13 @@ def compile_unix_full_gpu_debug(lib_name) {
     }]
 }
 
-def compile_unix_cmake_gpu(lib_name) {
-    return ['GPU: CMake': {
-      node(NODE_LINUX_CPU) {
-        ws('workspace/build-cmake-gpu') {
-          timeout(time: max_time, unit: 'MINUTES') {
-            utils.init_git()
-            utils.docker_run('ubuntu_gpu_cu101', 'build_ubuntu_gpu_cmake', false)
-            utils.pack_lib(lib_name, mx_cmake_lib_cython)
-          }
-        }
-      }
-    }]
-}
-
 def compile_unix_tensorrt_gpu(lib_name) {
     return ['TensorRT': {
       node(NODE_LINUX_CPU) {
         ws('workspace/build-tensorrt') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_gpu_cu102', 'build_ubuntu_gpu_tensorrt', false)
+            utils.docker_run('ubuntu_tensorrt_cu111', 'build_ubuntu_gpu_tensorrt', false)
             utils.pack_lib(lib_name, mx_tensorrt_lib)
           }
         }
@@ -510,7 +496,7 @@ def compile_unix_clang10_cuda_werror(lib_name) {
         ws('workspace/build-cpu-clang10') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.init_git()
-            utils.docker_run('ubuntu_gpu_cu101', 'build_ubuntu_gpu_clang10_werror', false)
+            utils.docker_run('ubuntu_gpu_cu111', 'build_ubuntu_gpu_clang10_werror', false)
             utils.pack_lib(lib_name, mx_lib)
           }
         }
@@ -696,7 +682,7 @@ def test_unix_python3_gpu(lib_name) {
         ws('workspace/ut-python3-gpu') {
           try {
             utils.unpack_and_init(lib_name, mx_lib_cython)
-            python3_gpu_ut_cython('ubuntu_gpu_cu101')
+            python3_gpu_ut_cython('ubuntu_gpu_cu111')
             utils.publish_test_coverage()
           } finally {
             utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_gpu.xml')
@@ -778,7 +764,7 @@ def test_unix_python3_mkldnn_gpu(lib_name) {
         ws('workspace/ut-python3-mkldnn-gpu') {
           try {
             utils.unpack_and_init(lib_name, mx_mkldnn_lib)
-            python3_gpu_ut('ubuntu_gpu_cu101')
+            python3_gpu_ut('ubuntu_gpu_cu111')
             utils.publish_test_coverage()
           } finally {
             utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_mkldnn_gpu.xml')
@@ -794,7 +780,7 @@ def test_unix_python3_mkldnn_nocudnn_gpu(lib_name) {
         ws('workspace/ut-python3-mkldnn-gpu-nocudnn') {
           try {
             utils.unpack_and_init(lib_name, mx_mkldnn_lib)
-            python3_gpu_ut_nocudnn('ubuntu_gpu_cu101')
+            python3_gpu_ut_nocudnn('ubuntu_gpu_cu111')
             utils.publish_test_coverage()
           } finally {
             utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_mkldnn_gpu_nocudnn.xml')
@@ -838,7 +824,7 @@ def test_unix_byteps_gpu(lib_name) {
         ws('workspace/it-byteps') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.unpack_and_init(lib_name, mx_lib)
-            utils.docker_run('ubuntu_gpu_cu101', 'integrationtest_ubuntu_gpu_byteps', true, '32768m')
+            utils.docker_run('ubuntu_gpu_cu111', 'integrationtest_ubuntu_gpu_byteps', true, '32768m')
             utils.publish_test_coverage()
           }
         }
@@ -852,7 +838,7 @@ def test_unix_distributed_kvstore_gpu(lib_name) {
         ws('workspace/it-dist-kvstore') {
           timeout(time: max_time, unit: 'MINUTES') {
             utils.unpack_and_init(lib_name, mx_lib)
-            utils.docker_run('ubuntu_gpu_cu101', 'integrationtest_ubuntu_gpu_dist_kvstore', true)
+            utils.docker_run('ubuntu_gpu_cu111', 'integrationtest_ubuntu_gpu_dist_kvstore', true)
             utils.publish_test_coverage()
           }
         }
diff --git a/ci/jenkins/Jenkinsfile_unix_gpu b/ci/jenkins/Jenkinsfile_unix_gpu
index 6fbdc751ea90..63bd3d29b641 100644
--- a/ci/jenkins/Jenkinsfile_unix_gpu
+++ b/ci/jenkins/Jenkinsfile_unix_gpu
@@ -38,7 +38,6 @@ core_logic: {
     custom_steps.compile_unix_mkldnn_nocudnn_gpu('mkldnn_gpu_nocudnn'),
     custom_steps.compile_unix_full_gpu('gpu'),
     custom_steps.compile_unix_full_gpu_debug('gpu_debug'),
-    custom_steps.compile_unix_cmake_gpu('cmake_gpu'),
     custom_steps.compile_unix_tensorrt_gpu('tensorrt'),
     custom_steps.compile_unix_int64_gpu('gpu_int64'),
   ])
diff --git a/python/mxnet/contrib/text/utils.py b/python/mxnet/contrib/text/utils.py
index 79cf903edbee..46e93daecb01 100644
--- a/python/mxnet/contrib/text/utils.py
+++ b/python/mxnet/contrib/text/utils.py
@@ -77,7 +77,7 @@ def count_tokens_from_str(source_str, token_delim=' ', seq_delim='\n',
         source_str = [t.lower() for t in source_str]
 
     if counter_to_update is None:
-        return collections.Counter(source_str)
+        return collections.Counter(source_str)  # pylint: disable=too-many-function-args
     else:
         counter_to_update.update(source_str)
         return counter_to_update
diff --git a/python/mxnet/numpy/multiarray.py b/python/mxnet/numpy/multiarray.py
index 6f74b49d01a2..e260a41849ad 100644
--- a/python/mxnet/numpy/multiarray.py
+++ b/python/mxnet/numpy/multiarray.py
@@ -260,8 +260,8 @@ def _wrap_mxnp_np_ufunc(x1, x2):
         return func(x1, x2)
     return _wrap_mxnp_np_ufunc
 
-@set_module('mxnet.numpy')  # pylint: disable=invalid-name
-class ndarray(NDArray):
+@set_module('mxnet.numpy')
+class ndarray(NDArray):  # pylint: disable=invalid-name
     """
     ndarray(handle, writable=True):
 
diff --git a/src/profiler/profiler.h b/src/profiler/profiler.h
index 132a9f90ec68..6dc3cf46c895 100644
--- a/src/profiler/profiler.h
+++ b/src/profiler/profiler.h
@@ -54,7 +54,12 @@ struct static_string {
   inline explicit static_string(const char *s) { set(s); }
   inline const char *c_str() const { return &string_[0]; }
   inline void set(const char *s) {
+#pragma GCC diagnostic push
+#if __GNUC__ >= 8
+#pragma GCC diagnostic ignored "-Wstringop-truncation"
+#endif
     strncpy(&string_[0], s, string_size - 1);
+#pragma GCC diagnostic pop
     string_[string_size - 1] = '\0';
   }
   inline void append(const char *s) {