diff --git a/benchmark/opperf/opperf.py b/benchmark/opperf/opperf.py
old mode 100755
new mode 100644
diff --git a/cd/python/docker/Dockerfile.test b/cd/python/docker/Dockerfile.test
index bed059d0fc73..6ded5f4722fb 100644
--- a/cd/python/docker/Dockerfile.test
+++ b/cd/python/docker/Dockerfile.test
@@ -24,7 +24,7 @@ ARG BASE_IMAGE
 FROM ${BASE_IMAGE}
 
 # Install test dependencies
-RUN pip install nose
+RUN pip install pytest
 
 ARG USER_ID=1001
 ARG GROUP_ID=1001
diff --git a/cd/python/pypi/pypi_publish.py b/cd/python/pypi/pypi_publish.py
old mode 100755
new mode 100644
diff --git a/cd/utils/artifact_repository.py b/cd/utils/artifact_repository.py
old mode 100755
new mode 100644
diff --git a/ci/README.md b/ci/README.md
index 155a0104a125..f5e847eac3bd 100644
--- a/ci/README.md
+++ b/ci/README.md
@@ -185,7 +185,7 @@ sudo pip3 install --upgrade --force-reinstall build/mxnet-1.3.1-py2.py3-none-any
 
 # Execute a single python test:
 
-nosetests-3.4 -v -s tests/python/unittest/test_ndarray.py
+pytest -v --capture=no tests/python/unittest/test_ndarray.py
 
 
 # Debug with cgdb
diff --git a/ci/dev_menu.py b/ci/dev_menu.py
old mode 100755
new mode 100644
index e9f031e1b171..b14f7f296c12
--- a/ci/dev_menu.py
+++ b/ci/dev_menu.py
@@ -119,7 +119,7 @@ def provision_virtualenv(venv_path=DEFAULT_PYENV):
         provision_virtualenv,
     ]),
     ('[Local] Python Unit tests',
-        "./py3_venv/bin/nosetests -v tests/python/unittest/"
+        "pytest -v tests/python/unittest/"
     ),
     ('[Docker] Build the MXNet binary - outputs to "lib/"',
         "ci/build.py --platform ubuntu_cpu_lite /work/runtime_functions.sh build_ubuntu_cpu_docs"),
diff --git a/ci/docker/Dockerfile.build.centos7_cpu b/ci/docker/Dockerfile.build.centos7_cpu
index 0cfa5a9f6e47..b16984706146 100644
--- a/ci/docker/Dockerfile.build.centos7_cpu
+++ b/ci/docker/Dockerfile.build.centos7_cpu
@@ -24,10 +24,14 @@ WORKDIR /work/deps
 
 COPY install/centos7_core.sh /work/
 RUN /work/centos7_core.sh
+
 COPY install/centos7_ccache.sh /work/
 RUN /work/centos7_ccache.sh
+
 COPY install/centos7_python.sh /work/
+COPY install/requirements /work/
 RUN /work/centos7_python.sh
+
 COPY install/centos7_scala.sh /work/
 RUN /work/centos7_scala.sh
 
diff --git a/ci/docker/Dockerfile.build.centos7_gpu b/ci/docker/Dockerfile.build.centos7_gpu
index 7e49e88b3a52..389a17eab12b 100644
--- a/ci/docker/Dockerfile.build.centos7_gpu
+++ b/ci/docker/Dockerfile.build.centos7_gpu
@@ -24,9 +24,12 @@ WORKDIR /work/deps
 
 COPY install/centos7_core.sh /work/
 RUN /work/centos7_core.sh
+
 COPY install/centos7_ccache.sh /work/
 RUN /work/centos7_ccache.sh
+
 COPY install/centos7_python.sh /work/
+COPY install/requirements /work/
 RUN /work/centos7_python.sh
 
 ENV CUDNN_VERSION=7.6.0.64
diff --git a/ci/docker/Dockerfile.build.ubuntu_cpu b/ci/docker/Dockerfile.build.ubuntu_cpu
index b1eb89bb3f36..266109c05cbd 100644
--- a/ci/docker/Dockerfile.build.ubuntu_cpu
+++ b/ci/docker/Dockerfile.build.ubuntu_cpu
@@ -61,9 +61,6 @@ RUN /work/ubuntu_mkl.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 RUN /work/ubuntu_docs.sh
 
diff --git a/ci/docker/Dockerfile.build.ubuntu_cpu_julia b/ci/docker/Dockerfile.build.ubuntu_cpu_julia
index b1eb89bb3f36..266109c05cbd 100644
--- a/ci/docker/Dockerfile.build.ubuntu_cpu_julia
+++ b/ci/docker/Dockerfile.build.ubuntu_cpu_julia
@@ -61,9 +61,6 @@ RUN /work/ubuntu_mkl.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 RUN /work/ubuntu_docs.sh
 
diff --git a/ci/docker/Dockerfile.build.ubuntu_cpu_python b/ci/docker/Dockerfile.build.ubuntu_cpu_python
index 6b217d4d341d..d1b0f6575100 100644
--- a/ci/docker/Dockerfile.build.ubuntu_cpu_python
+++ b/ci/docker/Dockerfile.build.ubuntu_cpu_python
@@ -32,9 +32,6 @@ COPY install/ubuntu_python.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_python.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 RUN /work/ubuntu_docs.sh
 
@@ -46,4 +43,4 @@ RUN /work/ubuntu_adduser.sh
 
 COPY runtime_functions.sh /work/
 
-WORKDIR /work/mxnet
\ No newline at end of file
+WORKDIR /work/mxnet
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu100 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu100
index e35c64eeca5d..2c4a34a0ff51 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu100
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu100
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu101 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu101
index aa62fbc6307e..fb124f4abff7 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu101
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu101
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu102 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu102
index 8badadbb1bdb..0fec5d8790b6 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu102
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu102
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu80 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu80
index 30971b0a5c6e..09996ec9b95e 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu80
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu80
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu90 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu90
index cc50e7e55191..910631c927c0 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu90
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu90
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_gpu_cu92 b/ci/docker/Dockerfile.build.ubuntu_gpu_cu92
index 40a4f44abeb5..481ae4260c32 100644
--- a/ci/docker/Dockerfile.build.ubuntu_gpu_cu92
+++ b/ci/docker/Dockerfile.build.ubuntu_gpu_cu92
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_nightly_cpu b/ci/docker/Dockerfile.build.ubuntu_nightly_cpu
index 5717df1b9130..e7d42557fd7e 100644
--- a/ci/docker/Dockerfile.build.ubuntu_nightly_cpu
+++ b/ci/docker/Dockerfile.build.ubuntu_nightly_cpu
@@ -49,9 +49,6 @@ RUN /work/ubuntu_clang.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/Dockerfile.build.ubuntu_nightly_gpu b/ci/docker/Dockerfile.build.ubuntu_nightly_gpu
index 5e812c433b43..75c5b37d4b12 100644
--- a/ci/docker/Dockerfile.build.ubuntu_nightly_gpu
+++ b/ci/docker/Dockerfile.build.ubuntu_nightly_gpu
@@ -55,9 +55,6 @@ RUN /work/ubuntu_llvm.sh
 COPY install/ubuntu_caffe.sh /work/
 RUN /work/ubuntu_caffe.sh
 
-COPY install/ubuntu_onnx.sh /work/
-RUN /work/ubuntu_onnx.sh
-
 COPY install/ubuntu_docs.sh /work/
 COPY install/requirements /work/
 RUN /work/ubuntu_docs.sh
diff --git a/ci/docker/install/centos7_python.sh b/ci/docker/install/centos7_python.sh
index 06c53bea48c1..35202f2bbb2d 100755
--- a/ci/docker/install/centos7_python.sh
+++ b/ci/docker/install/centos7_python.sh
@@ -29,4 +29,4 @@ yum -y install python36u
 # Install PIP
 curl "https://bootstrap.pypa.io/get-pip.py" -o "get-pip.py"
 python3.6 get-pip.py
-pip3 install nose pylint numpy nose-timer requests h5py scipy==1.2.3
+pip3 install -r /work/requirements
diff --git a/ci/docker/install/requirements b/ci/docker/install/requirements
index 2d5125e8e2da..fc3763494a5b 100644
--- a/ci/docker/install/requirements
+++ b/ci/docker/install/requirements
@@ -24,11 +24,17 @@ Cython==0.29.7
 decorator==4.4.0
 h5py==2.8.0rc1
 mock==2.0.0
-nose==1.3.7
-nose-timer==0.7.3
 numpy>1.16.0,<2.0.0
 pylint==2.3.1; python_version >= '3.0'
 requests<2.19.0,>=2.18.4
 scipy==1.2.1
 six==1.11.0
 setuptools
+pytest==5.3.2
+pytest-env==0.6.2
+pytest-cov==2.8.1
+pytest-xdist==1.31.0
+protobuf==3.5.2
+onnx==1.3.0
+Pillow==5.0.0
+tabulate==0.7.5
diff --git a/ci/docker/install/ubuntu_onnx.sh b/ci/docker/install/ubuntu_onnx.sh
deleted file mode 100755
index 44d6b9ed52dc..000000000000
--- a/ci/docker/install/ubuntu_onnx.sh
+++ /dev/null
@@ -1,34 +0,0 @@
-#!/usr/bin/env bash
-
-# 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.
-
-######################################################################
-# This script installs ONNX for Python along with all required dependencies 
-# on a Ubuntu Machine.
-# Tested on Ubuntu 16.04 distro.
-######################################################################
-
-set -e
-set -x
-
-echo "Installing libprotobuf-dev and protobuf-compiler ..."
-apt-get update || true
-apt-get install -y libprotobuf-dev protobuf-compiler
-
-echo "Installing pytest, pytest-cov, protobuf, Pillow, ONNX and tabulate ..."
-pip3 install pytest==3.6.3 pytest-cov==2.5.1 protobuf==3.5.2 onnx==1.3.0 Pillow==5.0.0 tabulate==0.7.5
diff --git a/ci/docker/install/ubuntu_publish.sh b/ci/docker/install/ubuntu_publish.sh
index 5b13e1306fbd..4d20cb2f4b78 100755
--- a/ci/docker/install/ubuntu_publish.sh
+++ b/ci/docker/install/ubuntu_publish.sh
@@ -77,7 +77,7 @@ apt-get install -y python python-pip python3 python3-pip
 python3 -m pip install --upgrade 'pip<19'
 
 # Restrict numpy version to <1.18 due to use of Python 3.4 on Ubuntu 14.04
-python3 -m pip install --upgrade --ignore-installed nose cpplint==1.3.0 pylint==2.3.1 'numpy>1.16.0,<1.18' nose-timer 'requests<2.19.0,>=2.18.4' h5py==2.8.0rc1 scipy==1.0.1 boto3
+python3 -m pip install --upgrade --ignore-installed cpplint==1.3.0 pylint==2.3.1 'numpy>1.16.0,<1.18' 'requests<2.19.0,>=2.18.4' h5py==2.8.0rc1 scipy==1.0.1 boto3 pytest==4.6.9 pytest-env==0.6.2 pytest-cov==2.8.1 pytest-xdist==1.31.0
 
 # CMake 3.13.2+ is required
 mkdir /opt/cmake && cd /opt/cmake
diff --git a/ci/docker/install/ubuntu_python.sh b/ci/docker/install/ubuntu_python.sh
index b6792a286fad..6234aac12283 100755
--- a/ci/docker/install/ubuntu_python.sh
+++ b/ci/docker/install/ubuntu_python.sh
@@ -24,6 +24,7 @@ set -ex
 # install libraries for mxnet's python package on ubuntu
 apt-get update || true
 apt-get install -y python-dev python3-dev virtualenv wget
+apt-get install -y libprotobuf-dev protobuf-compiler
 
 # the version of the pip shipped with ubuntu may be too lower, install a recent version here
 wget -nv https://bootstrap.pypa.io/get-pip.py
diff --git a/ci/docker/qemu/runtime_functions.py b/ci/docker/qemu/runtime_functions.py
index 5a57cb8dae6a..b8306bbaf19f 100755
--- a/ci/docker/qemu/runtime_functions.py
+++ b/ci/docker/qemu/runtime_functions.py
@@ -77,12 +77,10 @@ def run_ut_python3_qemu_internal():
     logging.info("=== NOW Running inside QEMU ===")
     logging.info("PIP Installing %s", pkg)
     check_call(['sudo', 'pip3', 'install', pkg])
-    logging.info("PIP Installing mxnet/test_requirements.txt") 
+    logging.info("PIP Installing mxnet/test_requirements.txt")
     check_call(['sudo', 'pip3', 'install', '-r', 'mxnet/test_requirements.txt'])
     logging.info("Running tests in mxnet/tests/python/unittest/")
-    check_call(['nosetests', '--with-timer', '--with-xunit', '--xunit-file', 'nosetests_unittest.xml', '--verbose', 'mxnet/tests/python/unittest/test_engine.py'])
-    # Example to run a single unit test:
-    # check_call(['nosetests', '--with-timer', '--with-xunit', '--xunit-file', 'nosetests_unittest.xml', '--verbose', 'mxnet/tests/python/unittest/test_ndarray.py:test_ndarray_fluent'])
+    check_call(['pytest', '--durations=50', '--cov-report', 'xml:tests_unittest.xml', '-v', 'mxnet/tests/python/unittest/test_engine.py'])
 
 
 
diff --git a/ci/docker/runtime_functions.sh b/ci/docker/runtime_functions.sh
index e171767d51f3..c39e3967371b 100755
--- a/ci/docker/runtime_functions.sh
+++ b/ci/docker/runtime_functions.sh
@@ -22,8 +22,6 @@
 
 set -ex
 
-NOSE_COVERAGE_ARGUMENTS="--with-coverage --cover-inclusive --cover-xml --cover-branches --cover-package=mxnet"
-NOSE_TIMER_ARGUMENTS="--with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error"
 CI_CUDA_COMPUTE_CAPABILITIES="-gencode=arch=compute_52,code=sm_52 -gencode=arch=compute_70,code=sm_70"
 CI_CMAKE_CUDA_ARCH="5.2 7.0"
 
@@ -968,7 +966,7 @@ sanity_check() {
     make cpplint jnilint
     make -f R-package/Makefile rcpplint
     make pylint
-    nosetests-3.4 tests/tutorials/test_sanity_tutorials.py
+    pytest tests/tutorials/test_sanity_tutorials.py
 }
 
 # Tests libmxnet
@@ -988,10 +986,8 @@ cd_unittest_ubuntu() {
     local mxnet_variant=${1:?"This function requires a mxnet variant as the first argument"}
     local python_cmd=${2:?"This function requires a python command as the first argument"}
 
-    local nose_cmd="nosetests-3.4"
-
-    $nose_cmd $NOSE_TIMER_ARGUMENTS --verbose tests/python/unittest
-    $nose_cmd $NOSE_TIMER_ARGUMENTS --verbose tests/python/quantization
+    pytest --durations=50 --verbose tests/python/unittest
+    pytest --durations=50 --verbose tests/python/quantization
 
     # https://github.com/apache/incubator-mxnet/issues/11801
     # if [[ ${mxnet_variant} = "cpu" ]] || [[ ${mxnet_variant} = "mkl" ]]; then
@@ -999,7 +995,7 @@ cd_unittest_ubuntu() {
     # fi
 
     if [[ ${mxnet_variant} = cu* ]]; then
-        $nose_cmd $NOSE_TIMER_ARGUMENTS --verbose tests/python/gpu
+        pytest --durations=50 --verbose tests/python/gpu
 
         # Adding these here as CI doesn't test all CUDA environments
         $python_cmd example/image-classification/test_score.py
@@ -1010,7 +1006,7 @@ cd_unittest_ubuntu() {
         # skipping python 2 testing
         # https://github.com/apache/incubator-mxnet/issues/14675
         if [[ ${python_cmd} = "python3" ]]; then
-            $nose_cmd $NOSE_TIMER_ARGUMENTS --verbose tests/python/mkl
+            pytest --durations=50 --verbose tests/python/mkl
         fi
     fi
 }
@@ -1023,8 +1019,8 @@ unittest_ubuntu_python3_cpu() {
     export MXNET_SUBGRAPH_VERBOSE=0
     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_unittest.xml --verbose tests/python/unittest
-    nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_quantization.xml --verbose tests/python/quantization
+    pytest --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
+    pytest --durations=50 --cov-report xml:tests_quantization.xml --verbose tests/python/quantization
 }
 
 unittest_ubuntu_python3_cpu_mkldnn() {
@@ -1035,8 +1031,8 @@ unittest_ubuntu_python3_cpu_mkldnn() {
     export MXNET_SUBGRAPH_VERBOSE=0
     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_unittest.xml --verbose tests/python/unittest
-    nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_mkl.xml --verbose tests/python/mkl
+    pytest --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
+    pytest --durations=50 --cov-report xml:tests_mkl.xml --verbose tests/python/mkl
 }
 
 unittest_ubuntu_python3_gpu() {
@@ -1048,7 +1044,7 @@ unittest_ubuntu_python3_gpu() {
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
     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_gpu.xml --verbose tests/python/gpu
+    pytest --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
 }
 
 unittest_ubuntu_python3_gpu_cython() {
@@ -1062,7 +1058,7 @@ unittest_ubuntu_python3_gpu_cython() {
     export MXNET_ENFORCE_CYTHON=1
     export DMLC_LOG_STACK_TRACE_DEPTH=10
     check_cython
-    nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_gpu.xml --verbose tests/python/gpu
+    pytest --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
 }
 
 unittest_ubuntu_python3_gpu_nocudnn() {
@@ -1073,7 +1069,7 @@ unittest_ubuntu_python3_gpu_nocudnn() {
     export CUDNN_OFF_TEST_ONLY=true
     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_gpu.xml --verbose tests/python/gpu
+    pytest --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
 }
 
 unittest_ubuntu_tensorrt_gpu() {
@@ -1086,7 +1082,7 @@ unittest_ubuntu_tensorrt_gpu() {
     export MXNET_ENABLE_CYTHON=0
     export DMLC_LOG_STACK_TRACE_DEPTH=10
     tests/python/tensorrt/lenet5_train.py
-    nosetests-3.4 $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_trt_gpu.xml --verbose --nocapture tests/python/tensorrt/
+    pytest --durations=50 --cov-report xml:tests_trt_gpu.xml --verbose --capture=no tests/python/tensorrt/
 }
 
 # quantization gpu currently only runs on P3 instances
@@ -1100,7 +1096,7 @@ unittest_ubuntu_python3_quantization_gpu() {
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
     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
+    pytest --durations=50 --cov-report xml:tests_quantization_gpu.xml --verbose tests/python/quantization_gpu
 }
 
 unittest_centos7_cpu_scala() {
@@ -1233,8 +1229,8 @@ unittest_ubuntu_cpu_julia10() {
 unittest_centos7_cpu() {
     set -ex
     cd /work/mxnet
-    python3.6 -m "nose" $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_unittest.xml --verbose tests/python/unittest
-    python3.6 -m "nose" $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_train.xml --verbose tests/python/train
+    python3.6 -m pytest --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
+    python3.6 -m pytest --durations=50 --cov-report xml:tests_train.xml --verbose tests/python/train
 }
 
 unittest_centos7_gpu() {
@@ -1242,17 +1238,17 @@ unittest_centos7_gpu() {
     cd /work/mxnet
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
     export DMLC_LOG_STACK_TRACE_DEPTH=10
-    python3.6 -m "nose" $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_gpu.xml --verbose tests/python/gpu
+    python3.6 -m pytest --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
 }
 
 integrationtest_ubuntu_cpu_onnx() {
 	set -ex
 	export PYTHONPATH=./python/
     export DMLC_LOG_STACK_TRACE_DEPTH=10
-	tests/python-pytest/onnx/backend_test.py
-	pytest tests/python-pytest/onnx/mxnet_export_test.py
-	pytest tests/python-pytest/onnx/test_models.py
-	pytest tests/python-pytest/onnx/test_node.py
+	pytest tests/python/unittest/onnx/backend_test.py
+	pytest tests/python/unittest/onnx/mxnet_export_test.py
+	pytest tests/python/unittest/onnx/test_models.py
+	pytest tests/python/unittest/onnx/test_node.py
 }
 
 integrationtest_ubuntu_gpu_python() {
@@ -1361,10 +1357,9 @@ test_ubuntu_cpu_python3() {
     source $VENV/bin/activate
 
     cd /work/mxnet/python
-    pip3 install nose nose-timer
     pip3 install -e .
     cd /work/mxnet
-    python3 -m "nose" $NOSE_COVERAGE_ARGUMENTS $NOSE_TIMER_ARGUMENTS --verbose tests/python/unittest
+    python3 -m pytest $TEST_TIMER_ARGUMENTS --verbose tests/python/unittest
 
     popd
 }
@@ -1440,9 +1435,9 @@ nightly_test_large_tensor() {
     set -ex
     export PYTHONPATH=./python/
     export DMLC_LOG_STACK_TRACE_DEPTH=10
-    nosetests-3.4 tests/nightly/test_large_array.py:test_tensor
-    nosetests-3.4 tests/nightly/test_large_array.py:test_nn
-    nosetests-3.4 tests/nightly/test_large_array.py:test_basic
+    pytest tests/nightly/test_large_array.py::test_tensor
+    pytest tests/nightly/test_large_array.py::test_nn
+    pytest tests/nightly/test_large_array.py::test_basic
 }
 
 #Test Large Vectors
@@ -1450,9 +1445,9 @@ nightly_test_large_vector() {
     set -ex
     export PYTHONPATH=./python/
     export DMLC_LOG_STACK_TRACE_DEPTH=10
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_tensor
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_nn
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_basic
+    pytest tests/nightly/test_large_vector.py::test_tensor
+    pytest tests/nightly/test_large_vector.py::test_nn
+    pytest tests/nightly/test_large_vector.py::test_basic
 }
 
 #Test Large Vectors
@@ -1460,9 +1455,9 @@ nightly_test_large_vector() {
     set -ex
     export PYTHONPATH=./python/
     export DMLC_LOG_STACK_TRACE_DEPTH=10
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_tensor
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_nn
-    nosetests-3.4 tests/nightly/test_large_vector.py:test_basic
+    pytest tests/nightly/test_large_vector.py::test_tensor
+    pytest tests/nightly/test_large_vector.py::test_nn
+    pytest tests/nightly/test_large_vector.py::test_basic
 }
 
 #Tests Amalgamation Build with 5 different sets of flags
@@ -1502,26 +1497,6 @@ nightly_model_backwards_compat_train() {
     ./tests/nightly/model_backwards_compatibility_check/train_mxnet_legacy_models.sh
 }
 
-nightly_straight_dope_python3_single_gpu_tests() {
-    set -ex
-    cd /work/mxnet/tests/nightly/straight_dope
-    export PYTHONPATH=/work/mxnet/python/
-    export MXNET_TEST_KERNEL=python3
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
-    nosetests-3.4 $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_straight_dope_python3_single_gpu.xml \
-      test_notebooks_single_gpu.py --nologcapture
-}
-
-nightly_straight_dope_python3_multi_gpu_tests() {
-    set -ex
-    cd /work/mxnet/tests/nightly/straight_dope
-    export PYTHONPATH=/work/mxnet/python/
-    export MXNET_TEST_KERNEL=python3
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
-    nosetests-3.4 $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_straight_dope_python3_multi_gpu.xml \
-      test_notebooks_multi_gpu.py --nologcapture
-}
-
 nightly_tutorial_test_ubuntu_python3_gpu() {
     set -ex
     cd /work/mxnet/docs
@@ -1534,7 +1509,7 @@ nightly_tutorial_test_ubuntu_python3_gpu() {
     export PYTHONPATH=/work/mxnet/python/
     export MXNET_TUTORIAL_TEST_KERNEL=python3
     cd /work/mxnet/tests/tutorials
-    nosetests-3.4 $NOSE_TIMER_ARGUMENTS --with-xunit --xunit-file nosetests_tutorials.xml test_tutorials.py --nologcapture
+    pytest --durations=50 --cov-report xml:tests_tutorials.xml --capture=no test_tutorials.py
 }
 
 nightly_java_demo_test_cpu() {
@@ -1558,8 +1533,8 @@ nightly_estimator() {
     export DMLC_LOG_STACK_TRACE_DEPTH=10
     cd /work/mxnet/tests/nightly/estimator
     export PYTHONPATH=/work/mxnet/python/
-    nosetests test_estimator_cnn.py
-    nosetests test_sentiment_rnn.py
+    pytest test_estimator_cnn.py
+    pytest test_sentiment_rnn.py
 }
 
 # For testing PRs
diff --git a/ci/docker_cache.py b/ci/docker_cache.py
old mode 100755
new mode 100644
diff --git a/ci/docker_login.py b/ci/docker_login.py
old mode 100755
new mode 100644
diff --git a/ci/jenkins/Jenkins_steps.groovy b/ci/jenkins/Jenkins_steps.groovy
index e5ce8de24485..96ff84dd6d55 100644
--- a/ci/jenkins/Jenkins_steps.groovy
+++ b/ci/jenkins/Jenkins_steps.groovy
@@ -808,8 +808,8 @@ def test_unix_python3_cpu() {
             python3_ut('ubuntu_cpu')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_cpu_unittest.xml')
-            utils.collect_test_results_unix('nosetests_quantization.xml', 'nosetests_python3_cpu_quantization.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_cpu_unittest.xml')
+            utils.collect_test_results_unix('tests_quantization.xml', 'tests_python3_cpu_quantization.xml')
           }
         }
       }
@@ -825,8 +825,8 @@ def test_unix_python3_mkl_cpu() {
             python3_ut('ubuntu_cpu')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_cpu_unittest.xml')
-            utils.collect_test_results_unix('nosetests_quantization.xml', 'nosetests_python3_cpu_quantization.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_cpu_unittest.xml')
+            utils.collect_test_results_unix('tests_quantization.xml', 'tests_python3_cpu_quantization.xml')
           }
         }
       }
@@ -842,7 +842,7 @@ def test_unix_python3_gpu() {
             python3_gpu_ut_cython('ubuntu_gpu_cu101')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_gpu.xml')
+            utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_gpu.xml')
           }
         }
       }
@@ -858,7 +858,7 @@ def test_unix_python3_gpu_no_tvm_op() {
             python3_gpu_ut_cython('ubuntu_gpu_cu101')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_gpu.xml')
+            utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_gpu.xml')
           }
         }
       }
@@ -875,7 +875,7 @@ def test_unix_python3_quantize_gpu() {
               utils.docker_run('ubuntu_gpu_cu101', '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')
+              utils.collect_test_results_unix('tests_quantization_gpu.xml', 'tests_python3_quantize_gpu.xml')
             }
           }
         }
@@ -891,8 +891,8 @@ def test_unix_python3_debug_cpu() {
             utils.unpack_and_init('cpu_debug', mx_cmake_lib_debug, true)
             python3_ut('ubuntu_cpu')
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_cpu_debug_unittest.xml')
-            utils.collect_test_results_unix('nosetests_quantization.xml', 'nosetests_python3_cpu_debug_quantization.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_cpu_debug_unittest.xml')
+            utils.collect_test_results_unix('tests_quantization.xml', 'tests_python3_cpu_debug_quantization.xml')
           }
         }
       }
@@ -907,8 +907,8 @@ def test_unix_python3_cpu_no_tvm_op() {
             utils.unpack_and_init('cpu_openblas_no_tvm_op', mx_cmake_lib_no_tvm_op)
             python3_ut('ubuntu_cpu')
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_cpu_no_tvm_op_unittest.xml')
-            utils.collect_test_results_unix('nosetests_quantization.xml', 'nosetests_python3_cpu_no_tvm_op_quantization.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_cpu_no_tvm_op_unittest.xml')
+            utils.collect_test_results_unix('tests_quantization.xml', 'tests_python3_cpu_no_tvm_op_quantization.xml')
           }
         }
       }
@@ -924,8 +924,8 @@ def test_unix_python3_mkldnn_cpu() {
             python3_ut_mkldnn('ubuntu_cpu')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_mkldnn_cpu_unittest.xml')
-            utils.collect_test_results_unix('nosetests_mkl.xml', 'nosetests_python3_mkldnn_cpu_mkl.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_mkldnn_cpu_unittest.xml')
+            utils.collect_test_results_unix('tests_mkl.xml', 'tests_python3_mkldnn_cpu_mkl.xml')
           }
         }
       }
@@ -941,8 +941,8 @@ def test_unix_python3_mkldnn_mkl_cpu() {
             python3_ut_mkldnn('ubuntu_cpu')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_mkldnn_cpu_unittest.xml')
-            utils.collect_test_results_unix('nosetests_mkl.xml', 'nosetests_python3_mkldnn_cpu_mkl.xml')
+            utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_mkldnn_cpu_unittest.xml')
+            utils.collect_test_results_unix('tests_mkl.xml', 'tests_python3_mkldnn_cpu_mkl.xml')
           }
         }
       }
@@ -958,7 +958,7 @@ def test_unix_python3_mkldnn_gpu() {
             python3_gpu_ut('ubuntu_gpu_cu101')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_mkldnn_gpu.xml')
+            utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_mkldnn_gpu.xml')
           }
         }
       }
@@ -974,7 +974,7 @@ def test_unix_python3_mkldnn_nocudnn_gpu() {
             python3_gpu_ut_nocudnn('ubuntu_gpu_cu101')
             utils.publish_test_coverage()
           } finally {
-            utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_mkldnn_gpu_nocudnn.xml')
+            utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_mkldnn_gpu_nocudnn.xml')
           }
         }
       }
@@ -991,7 +991,7 @@ def test_unix_python3_tensorrt_gpu() {
               utils.docker_run('ubuntu_gpu_tensorrt', 'unittest_ubuntu_tensorrt_gpu', true)
               utils.publish_test_coverage()
             } finally {
-              utils.collect_test_results_unix('nosetests_tensorrt.xml', 'nosetests_python3_tensorrt_gpu.xml')
+              utils.collect_test_results_unix('tests_tensorrt.xml', 'tests_python3_tensorrt_gpu.xml')
             }
           }
         }
@@ -1301,8 +1301,8 @@ def test_centos7_python3_cpu() {
               utils.docker_run('centos7_cpu', 'unittest_centos7_cpu', false)
               utils.publish_test_coverage()
             } finally {
-              utils.collect_test_results_unix('nosetests_unittest.xml', 'nosetests_python3_centos7_cpu_unittest.xml')
-              utils.collect_test_results_unix('nosetests_train.xml', 'nosetests_python3_centos7_cpu_train.xml')
+              utils.collect_test_results_unix('tests_unittest.xml', 'tests_python3_centos7_cpu_unittest.xml')
+              utils.collect_test_results_unix('tests_train.xml', 'tests_python3_centos7_cpu_train.xml')
             }
           }
         }
@@ -1320,7 +1320,7 @@ def test_centos7_python3_gpu() {
               utils.docker_run('centos7_gpu', 'unittest_centos7_gpu', true)
               utils.publish_test_coverage()
             } finally {
-              utils.collect_test_results_unix('nosetests_gpu.xml', 'nosetests_python3_centos7_gpu.xml')
+              utils.collect_test_results_unix('tests_gpu.xml', 'tests_python3_centos7_gpu.xml')
             }
           }
         }
@@ -1352,8 +1352,8 @@ def test_windows_python3_gpu() {
               unstash 'windows_package_gpu'
               powershell 'ci/windows/test_py3_gpu.ps1'
             } finally {
-              utils.collect_test_results_windows('nosetests_forward.xml', 'nosetests_gpu_forward_windows_python3_gpu.xml')
-              utils.collect_test_results_windows('nosetests_operator.xml', 'nosetests_gpu_operator_windows_python3_gpu.xml')
+              utils.collect_test_results_windows('tests_forward.xml', 'tests_gpu_forward_windows_python3_gpu.xml')
+              utils.collect_test_results_windows('tests_operator.xml', 'tests_gpu_operator_windows_python3_gpu.xml')
             }
           }
         }
@@ -1371,8 +1371,8 @@ def test_windows_python3_gpu_mkldnn() {
               unstash 'windows_package_gpu_mkldnn'
               powershell 'ci/windows/test_py3_gpu.ps1'
             } finally {
-              utils.collect_test_results_windows('nosetests_forward.xml', 'nosetests_gpu_forward_windows_python3_gpu_mkldnn.xml')
-              utils.collect_test_results_windows('nosetests_operator.xml', 'nosetests_gpu_operator_windows_python3_gpu_mkldnn.xml')
+              utils.collect_test_results_windows('tests_forward.xml', 'tests_gpu_forward_windows_python3_gpu_mkldnn.xml')
+              utils.collect_test_results_windows('tests_operator.xml', 'tests_gpu_operator_windows_python3_gpu_mkldnn.xml')
             }
           }
         }
@@ -1390,7 +1390,7 @@ def test_windows_python3_cpu() {
               unstash 'windows_package_cpu'
               powershell 'ci/windows/test_py3_cpu.ps1'
             } finally {
-              utils.collect_test_results_windows('nosetests_unittest.xml', 'nosetests_unittest_windows_python3_cpu.xml')
+              utils.collect_test_results_windows('tests_unittest.xml', 'tests_unittest_windows_python3_cpu.xml')
             }
           }
         }
diff --git a/ci/qemu/mxnet_requirements.txt b/ci/qemu/mxnet_requirements.txt
index 2ab0fd9612e5..48b87b3773ca 100644
--- a/ci/qemu/mxnet_requirements.txt
+++ b/ci/qemu/mxnet_requirements.txt
@@ -3,5 +3,7 @@ requests<2.19.0,>=2.18.4
 graphviz<0.9.0,>=0.8.1
 numpy>1.16.0,<2.0.0
 mock
-nose
-nose-timer
+pytest==5.3.2
+pytest-env==0.6.2
+pytest-cov==2.8.1
+pytest-xdist==1.31.0
diff --git a/ci/qemu/test_requirements.txt b/ci/qemu/test_requirements.txt
index 77037d89c673..1ea1a2e6ce56 100644
--- a/ci/qemu/test_requirements.txt
+++ b/ci/qemu/test_requirements.txt
@@ -1,3 +1,5 @@
 mock
-nose
-nose-timer
\ No newline at end of file
+pytest==5.3.2
+pytest-env==0.6.2
+pytest-cov==2.8.1
+pytest-xdist==1.31.0
diff --git a/ci/safe_docker_run.py b/ci/safe_docker_run.py
old mode 100755
new mode 100644
diff --git a/ci/test_docker_cache.py b/ci/test_docker_cache.py
index aeb399ff6b45..81b315be4cff 100644
--- a/ci/test_docker_cache.py
+++ b/ci/test_docker_cache.py
@@ -270,8 +270,3 @@ def _assert_docker_build(lambda_func, expected_cache_hit_count: int, expected_ca
         assert output.count('Using cache') == expected_cache_hit_count, \
             'Expected {} "Using cache", got {}. Log:{}'.\
                 format(expected_cache_hit_count, output.count('Using cache'), output)
-
-
-if __name__ == '__main__':
-    import nose
-    nose.main()
diff --git a/ci/test_docker_login.py b/ci/test_docker_login.py
index 6c989ade92ff..488295d9fa71 100644
--- a/ci/test_docker_login.py
+++ b/ci/test_docker_login.py
@@ -228,7 +228,3 @@ def test_main_default_argument_values(self, mock_login):
                 with self.assertRaises(RuntimeError):
                     main(["--secret-name", "name"])
 
-
-if __name__ == '__main__':
-    import nose
-    nose.main()
diff --git a/ci/test_safe_docker_run.py b/ci/test_safe_docker_run.py
index 433d42e8b2ea..5b75c0eef4ba 100644
--- a/ci/test_safe_docker_run.py
+++ b/ci/test_safe_docker_run.py
@@ -420,8 +420,3 @@ def wait_for_container(name: str) -> bool:
 
             # The container should no longer exist
             assert get_container(container_name) is None
-
-
-if __name__ == '__main__':
-    import nose
-    nose.main()
diff --git a/ci/windows/test_py3_cpu.ps1 b/ci/windows/test_py3_cpu.ps1
index 8f520bb5b259..6dc1e19fd4c8 100644
--- a/ci/windows/test_py3_cpu.ps1
+++ b/ci/windows/test_py3_cpu.ps1
@@ -24,11 +24,11 @@ $env:MXNET_SUBGRAPH_VERBOSE=0
 $env:MXNET_HOME=[io.path]::combine($PSScriptRoot, 'mxnet_home')
 
 C:\Python37\Scripts\pip install -r tests\requirements.txt
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_unittest.xml tests\python\unittest
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_unittest.xml tests\python\unittest
 if ($LastExitCode -ne 0) { Throw ("Error running unittest, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_train.xml tests\python\train
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_train.xml tests\python\train
 if ($LastExitCode -ne 0) { Throw ("Error running train tests, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
 # Adding this extra test since it's not possible to set env var on the fly in Windows.
 $env:MXNET_SAFE_ACCUMULATION=1
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_unittest.xml tests\python\unittest\test_operator.py:test_norm
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_unittest.xml tests\python\unittest\test_operator.py:test_norm
 if ($LastExitCode -ne 0) { Throw ("Error running unittest, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
diff --git a/ci/windows/test_py3_gpu.ps1 b/ci/windows/test_py3_gpu.ps1
index 0ce3d953d486..3cc4d18cf72c 100644
--- a/ci/windows/test_py3_gpu.ps1
+++ b/ci/windows/test_py3_gpu.ps1
@@ -24,15 +24,15 @@ $env:MXNET_SUBGRAPH_VERBOSE=0
 $env:MXNET_HOME=[io.path]::combine($PSScriptRoot, 'mxnet_home')
 
 C:\Python37\Scripts\pip install -r tests\requirements.txt
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_unittest.xml tests\python\unittest
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_unittest.xml tests\python\unittest
 if ($LastExitCode -ne 0) { Throw ("Error running unittest, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_operator.xml tests\python\gpu\test_operator_gpu.py
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_operator.xml tests\python\gpu\test_operator_gpu.py
 if ($LastExitCode -ne 0) { Throw ("Error running tests, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_forward.xml tests\python\gpu\test_forward.py
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_forward.xml tests\python\gpu\test_forward.py
 if ($LastExitCode -ne 0) { Throw ("Error running tests, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_train.xml tests\python\train
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_train.xml tests\python\train
 if ($LastExitCode -ne 0) { Throw ("Error running tests, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
 # Adding this extra test since it's not possible to set env var on the fly in Windows.
 $env:MXNET_SAFE_ACCUMULATION=1
-C:\Python37\python.exe -m nose -v --with-timer --timer-ok 1 --timer-warning 15 --timer-filter warning,error --with-xunit --xunit-file nosetests_operator.xml tests\python\gpu\test_operator_gpu.py:test_norm
+C:\Python37\python.exe -m pytest -v --durations=50 --cov-report xml:tests_operator.xml tests\python\gpu\test_operator_gpu.py:test_norm
 if ($LastExitCode -ne 0) { Throw ("Error running tests, python exited with status code " + ('{0:X}' -f $LastExitCode)) }
diff --git a/conftest.py b/conftest.py
new file mode 100644
index 000000000000..aedf2eb6ff36
--- /dev/null
+++ b/conftest.py
@@ -0,0 +1,221 @@
+# 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.
+"""conftest.py contains configuration for pytest.
+
+Configuration file for tests in tests/ and scripts/ folders.
+
+Note that fixtures of higher-scoped fixtures (such as ``session``) are
+instantiated before lower-scoped fixtures (such as ``function``).
+
+"""
+
+import logging
+import os
+import random
+
+import pytest
+
+
+def pytest_sessionfinish(session, exitstatus):
+    if exitstatus == 5:  # Don't fail if no tests were run
+        session.exitstatus = 0
+
+
+# * Random seed setup
+def pytest_configure():
+    """Pytest configuration hook to help reproduce test segfaults
+
+    Sets and outputs rng seeds.
+
+    The segfault-debug procedure on a module called test_module.py is:
+
+    1. run "pytest --verbose test_module.py".  A seg-faulting output might be:
+
+       [INFO] np, mx and python random seeds = 4018804151
+       test_module.test1 ... ok
+       test_module.test2 ... Illegal instruction (core dumped)
+
+    2. Copy the module-starting seed into the next command, then run:
+
+       MXNET_MODULE_SEED=4018804151 pytest --log-level=DEBUG --verbose test_module.py
+
+       Output might be:
+
+       [WARNING] **** module-level seed is set: all tests running deterministically ****
+       [INFO] np, mx and python random seeds = 4018804151
+       test_module.test1 ... [DEBUG] np and mx random seeds = 3935862516
+       ok
+       test_module.test2 ... [DEBUG] np and mx random seeds = 1435005594
+       Illegal instruction (core dumped)
+
+    3. Copy the segfaulting-test seed into the command:
+       MXNET_TEST_SEED=1435005594 pytest --log-level=DEBUG --verbose test_module.py:test2
+       Output might be:
+
+       [INFO] np, mx and python random seeds = 2481884723
+       test_module.test2 ... [DEBUG] np and mx random seeds = 1435005594
+       Illegal instruction (core dumped)
+
+    3. Finally reproduce the segfault directly under gdb (might need additional os packages)
+       by editing the bottom of test_module.py to be
+
+       if __name__ == '__main__':
+           logging.getLogger().setLevel(logging.DEBUG)
+           test2()
+
+       MXNET_TEST_SEED=1435005594 gdb -ex r --args python test_module.py
+
+    4. When finished debugging the segfault, remember to unset any exported MXNET_ seed
+       variables in the environment to return to non-deterministic testing (a good thing).
+    """
+
+    try:
+        import numpy as np
+        import mxnet as mx
+    except:
+        print('Unable to import numpy/mxnet. Skipping conftest.')
+        return
+    module_seed_str = os.getenv('MXNET_MODULE_SEED')
+    if module_seed_str is None:
+        seed = np.random.randint(0, np.iinfo(np.int32).max)
+    else:
+        seed = int(module_seed_str)
+        logging.warning('*** module-level seed is set: '
+                        'all tests running deterministically ***')
+    print('Setting module np/mx/python random seeds, '
+          'use MXNET_MODULE_SEED={} to reproduce.'.format(seed))
+
+    np.random.seed(seed)
+    mx.random.seed(seed)
+    random.seed(seed)
+
+    # The MXNET_TEST_SEED environment variable will override MXNET_MODULE_SEED for tests with
+    #  the 'with_seed()' decoration.  Inform the user of this once here at the module level.
+    if os.getenv('MXNET_TEST_SEED') is not None:
+        logging.warning('*** test-level seed set: all "@with_seed()" '
+                        'tests run deterministically ***')
+
+
+@pytest.hookimpl(tryfirst=True, hookwrapper=True)
+def pytest_runtest_makereport(item, call):
+    """Make test outcome available to fixture.
+
+    https://docs.pytest.org/en/latest/example/simple.html#making-test-result-information-available-in-fixtures
+    """
+    # execute all other hooks to obtain the report object
+    outcome = yield
+    rep = outcome.get_result()
+
+    # set a report attribute for each phase of a call, which can
+    # be "setup", "call", "teardown"
+    setattr(item, "rep_" + rep.when, rep)
+
+
+@pytest.fixture(scope='function', autouse=True)
+def function_scope_seed(request):
+    """A function scope fixture that manages rng seeds.
+
+    This fixture automatically initializes the python, numpy and mxnet random
+    number generators randomly on every test run.
+
+    def test_ok_with_random_data():
+        ...
+
+    To fix the seed used for a test case mark the test function with the
+    desired seed:
+
+    @pytest.mark.seed(1)
+    def test_not_ok_with_random_data():
+        '''This testcase actually works.'''
+        assert 17 == random.randint(0, 100)
+
+    When a test fails, the fixture outputs the seed used. The user can then set
+    the environment variable MXNET_TEST_SEED to the value reported, then rerun
+    the test with:
+
+        pytest --verbose -s <test_module_name.py> -k <failing_test>
+
+    To run a test repeatedly, install pytest-repeat and add the --count argument:
+
+        pip install pytest-repeat
+        pytest --verbose -s <test_module_name.py> -k <failing_test> --count 1000
+
+    """
+    try:
+        import numpy as np
+        import mxnet as mx
+    except:
+        print('Unable to import numpy/mxnet. Skipping conftest.')
+        return
+
+    seed = request.node.get_closest_marker('seed')
+    env_seed_str = os.getenv('MXNET_TEST_SEED')
+
+    if seed is not None:
+        seed = seed.args[0]
+        assert isinstance(seed, int)
+    elif env_seed_str is not None:
+        seed = int(env_seed_str)
+    else:
+        seed = np.random.randint(0, np.iinfo(np.int32).max)
+
+    post_test_state = np.random.get_state()
+    np.random.seed(seed)
+    mx.random.seed(seed)
+    random.seed(seed)
+
+    seed_message = ('np/mx/python random seeds are set to '
+                    '{}, use MXNET_TEST_SEED={} to reproduce.')
+    seed_message = seed_message.format(seed, seed)
+
+    # Always log seed on DEBUG log level. This makes sure we can find out the
+    # value of the seed even if the test case causes a segfault and subsequent
+    # teardown code is not run.
+    logging.debug(seed_message)
+
+    yield  # run the test
+
+    if request.node.rep_setup.failed:
+        logging.info("Setting up a test failed: {}", request.node.nodeid)
+    elif request.node.rep_call.outcome == 'failed':
+        # Either request.node.rep_setup.failed or request.node.rep_setup.passed
+        # should be True
+        assert request.node.rep_setup.passed
+        # On failure also log seed on INFO log level
+        logging.info(seed_message)
+
+    np.random.set_state(post_test_state)
+
+
+# * Shared test fixtures
+@pytest.fixture(params=[True, False])
+def hybridize(request):
+    return request.param
+
+@pytest.fixture(autouse=True)
+def doctest(doctest_namespace):
+    try:
+        import numpy as np
+        import mxnet as mx
+    except:
+        print('Unable to import numpy/mxnet. Skipping conftest.')
+        return
+    doctest_namespace['np'] = np
+    doctest_namespace['mx'] = mx
+    doctest_namespace['gluon'] = mx.gluon
+    import doctest
+    doctest.ELLIPSIS_MARKER = '-etc-'
diff --git a/contrib/clojure-package/examples/captcha/gen_captcha.py b/contrib/clojure-package/examples/captcha/gen_captcha.py
old mode 100755
new mode 100644
diff --git a/docker/install/python.sh b/docker/install/python.sh
index ba71246babbf..a8d0fd9217c9 100755
--- a/docker/install/python.sh
+++ b/docker/install/python.sh
@@ -24,5 +24,4 @@ apt-get update && apt-get install -y python-dev python3-dev
 # the version of the pip shipped with ubuntu may be too lower, install a recent version here
 cd /tmp && wget https://bootstrap.pypa.io/get-pip.py && python3 get-pip.py && python2 get-pip.py
 
-pip2 install nose pylint numpy nose-timer requests Pillow
-pip3 install nose pylint numpy nose-timer requests Pillow
+pip3 install pylint numpy requests Pillow pytest==5.3.2 pytest-env==0.6.2 pytest-cov==2.8.1 pytest-xdist==1.31.0
diff --git a/docs/static_site/src/pages/get_started/build_from_source.md b/docs/static_site/src/pages/get_started/build_from_source.md
index 1dfa95a82ade..ae391e4fe67e 100644
--- a/docs/static_site/src/pages/get_started/build_from_source.md
+++ b/docs/static_site/src/pages/get_started/build_from_source.md
@@ -221,7 +221,7 @@ make -j"$(nproc)"
 ```
 - Run test_nccl.py script as follows. The test should complete. It does not produce any output.
 ``` bash
-nosetests --verbose tests/python/gpu/test_nccl.py
+pytest --verbose tests/python/gpu/test_nccl.py
 ```
 
 **Recommendation to get the best performance out of NCCL:**
diff --git a/example/image-classification/__init__.py b/example/image-classification/__init__.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/benchmark.py b/example/image-classification/benchmark.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/benchmark_score.py b/example/image-classification/benchmark_score.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/common/data.py b/example/image-classification/common/data.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/common/fit.py b/example/image-classification/common/fit.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/fine-tune.py b/example/image-classification/fine-tune.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/score.py b/example/image-classification/score.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/symbols/alexnet.py b/example/image-classification/symbols/alexnet.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/symbols/resnet-v1.py b/example/image-classification/symbols/resnet-v1.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/symbols/resnetv1.py b/example/image-classification/symbols/resnetv1.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/test_score.py b/example/image-classification/test_score.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/train_cifar10.py b/example/image-classification/train_cifar10.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/train_imagenet.py b/example/image-classification/train_imagenet.py
old mode 100755
new mode 100644
diff --git a/example/image-classification/train_mnist.py b/example/image-classification/train_mnist.py
old mode 100755
new mode 100644
diff --git a/example/reinforcement-learning/dqn/dqn_demo.py b/example/reinforcement-learning/dqn/dqn_demo.py
old mode 100755
new mode 100644
diff --git a/example/reinforcement-learning/dqn/dqn_run_test.py b/example/reinforcement-learning/dqn/dqn_run_test.py
old mode 100755
new mode 100644
diff --git a/example/ssd/data/demo/download_demo_images.py b/example/ssd/data/demo/download_demo_images.py
old mode 100755
new mode 100644
diff --git a/example/ssd/dataset/pycocotools/__init__.py b/example/ssd/dataset/pycocotools/__init__.py
old mode 100755
new mode 100644
diff --git a/example/ssd/dataset/pycocotools/coco.py b/example/ssd/dataset/pycocotools/coco.py
old mode 100755
new mode 100644
diff --git a/example/ssd/demo.py b/example/ssd/demo.py
old mode 100755
new mode 100644
diff --git a/example/ssd/tools/prepare_dataset.py b/example/ssd/tools/prepare_dataset.py
old mode 100755
new mode 100644
diff --git a/example/ssd/train.py b/example/ssd/train.py
old mode 100755
new mode 100644
diff --git a/python/README.md b/python/README.md
index 148188537f48..e5d2faa86575 100644
--- a/python/README.md
+++ b/python/README.md
@@ -25,14 +25,14 @@ To install MXNet Python package, visit MXNet [Install Instruction](https://mxnet
 
 ## Running the unit tests
 
-For running unit tests, you will need the [nose PyPi package](https://pypi.python.org/pypi/nose). To install:
+For running unit tests, you will need the [pytest PyPi package](https://pypi.python.org/pypi/pytest). To install:
 ```bash
-pip install --upgrade nose
+pip install --upgrade pytest
 ```
 
-Once ```nose``` is installed, run the following from MXNet root directory (please make sure the installation path of ```nosetests``` is included in your ```$PATH``` environment variable):
+Once ```pytest``` is installed, run the following from MXNet root directory (please make sure the installation path of ```pytest``` is included in your ```$PATH``` environment variable):
 ```
-nosetests tests/python/unittest
-nosetests tests/python/train
+pytest tests/python/unittest
+pytest tests/python/train
 
 ```
diff --git a/python/mxnet/contrib/amp/amp.py b/python/mxnet/contrib/amp/amp.py
old mode 100755
new mode 100644
diff --git a/python/mxnet/contrib/amp/loss_scaler.py b/python/mxnet/contrib/amp/loss_scaler.py
old mode 100755
new mode 100644
diff --git a/python/mxnet/initializer.py b/python/mxnet/initializer.py
old mode 100755
new mode 100644
diff --git a/python/mxnet/module/executor_group.py b/python/mxnet/module/executor_group.py
old mode 100755
new mode 100644
diff --git a/python/mxnet/optimizer/optimizer.py b/python/mxnet/optimizer/optimizer.py
old mode 100755
new mode 100644
diff --git a/python/mxnet/test_utils.py b/python/mxnet/test_utils.py
old mode 100755
new mode 100644
diff --git a/tests/README.md b/tests/README.md
index de5d8107a790..b59335ea1593 100644
--- a/tests/README.md
+++ b/tests/README.md
@@ -68,7 +68,7 @@ Ninja is a build tool (like make) that prioritizes building speed. If you will b
     ```
 An example for running python tests would be
 ```
-ci/build.py --platform build_ubuntu_cpu_mkldnn /work/runtime_functions.sh unittest_ubuntu_python3_cpu PYTHONPATH=./python/ nosetests-2.7 tests/python/unittest
+ci/build.py --platform build_ubuntu_cpu_mkldnn /work/runtime_functions.sh unittest_ubuntu_python3_cpu PYTHONPATH=./python/ pytest tests/python/unittest
 ```
 
 
diff --git a/tests/jenkins/run_test.sh b/tests/jenkins/run_test.sh
index 5ded74291f25..48bb4da53fc4 100755
--- a/tests/jenkins/run_test.sh
+++ b/tests/jenkins/run_test.sh
@@ -42,16 +42,10 @@ export MXNET_ENGINE_INFO=false
 export PYTHONPATH=$(pwd)/python
 
 echo "BUILD python_test"
-nosetests --verbose tests/python/unittest || exit -1
-nosetests --verbose tests/python/gpu/test_operator_gpu.py || exit -1
-nosetests --verbose tests/python/gpu/test_forward.py || exit -1
-nosetests --verbose tests/python/train || exit -1
-
-echo "BUILD python3_test"
-nosetests3 --verbose tests/python/unittest || exit -1
-nosetests3 --verbose tests/python/gpu/test_operator_gpu.py || exit -1
-nosetests3 --verbose tests/python/gpu/test_forward.py || exit -1
-nosetests3 --verbose tests/python/train || exit -1
+pytest --verbose tests/python/unittest || exit -1
+pytest --verbose tests/python/gpu/test_operator_gpu.py || exit -1
+pytest --verbose tests/python/gpu/test_forward.py || exit -1
+pytest --verbose tests/python/train || exit -1
 
 echo "BUILD scala_test"
 export PATH=$PATH:/opt/apache-maven/bin
diff --git a/tests/jenkins/run_test_amzn_linux_gpu.sh b/tests/jenkins/run_test_amzn_linux_gpu.sh
index 57d9c7884088..a257b9684ba0 100755
--- a/tests/jenkins/run_test_amzn_linux_gpu.sh
+++ b/tests/jenkins/run_test_amzn_linux_gpu.sh
@@ -53,12 +53,8 @@ export MXNET_ENGINE_INFO=false
 export PYTHONPATH=${PWD}/python
 
 echo "BUILD python_test"
-nosetests --verbose tests/python/unittest
-nosetests --verbose tests/python/train
-
-echo "BUILD python3_test"
-nosetests3 --verbose tests/python/unittest
-nosetests3 --verbose tests/python/train
+pytest --verbose tests/python/unittest
+pytest --verbose tests/python/train
 
 #echo "BUILD julia_test"
 #export MXNET_HOME="${PWD}"
diff --git a/tests/jenkins/run_test_ubuntu.sh b/tests/jenkins/run_test_ubuntu.sh
index 0459d2cc8ec5..9c3d3c55c852 100755
--- a/tests/jenkins/run_test_ubuntu.sh
+++ b/tests/jenkins/run_test_ubuntu.sh
@@ -54,16 +54,10 @@ make -j$(nproc)
 export PYTHONPATH=${PWD}/python
 
 echo "BUILD python_test"
-nosetests --verbose tests/python/unittest || exit 1
-nosetests --verbose tests/python/gpu/test_operator_gpu.py || exit 1
-nosetests --verbose tests/python/gpu/test_forward.py || exit 1
-nosetests --verbose tests/python/train || exit 1
-
-echo "BUILD python3_test"
-nosetests3 --verbose tests/python/unittest || exit 1
-nosetests3 --verbose tests/python/gpu/test_operator_gpu.py || exit 1
-nosetests3 --verbose tests/python/gpu/test_forward.py || exit 1
-nosetests3 --verbose tests/python/train || exit 1
+pytest --verbose tests/python/unittest || exit 1
+pytest --verbose tests/python/gpu/test_operator_gpu.py || exit 1
+pytest --verbose tests/python/gpu/test_forward.py || exit 1
+pytest --verbose tests/python/train || exit 1
 
 echo "BUILD scala_test"
 export PATH=$PATH:/opt/apache-maven/bin
diff --git a/tests/nightly/broken_link_checker_test/test_broken_links.py b/tests/nightly/broken_link_checker_test/test_broken_links.py
old mode 100755
new mode 100644
diff --git a/tests/nightly/compilation_warnings/process_output.py b/tests/nightly/compilation_warnings/process_output.py
old mode 100755
new mode 100644
diff --git a/tests/nightly/estimator/test_estimator_cnn.py b/tests/nightly/estimator/test_estimator_cnn.py
index af519536dbed..0d113cdf4984 100644
--- a/tests/nightly/estimator/test_estimator_cnn.py
+++ b/tests/nightly/estimator/test_estimator_cnn.py
@@ -155,7 +155,3 @@ def test_estimator_gpu():
 
     assert acc.get()[1] > 0.80
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/nightly/estimator/test_sentiment_rnn.py b/tests/nightly/estimator/test_sentiment_rnn.py
index ab124ba95db3..367c69b88a0b 100644
--- a/tests/nightly/estimator/test_sentiment_rnn.py
+++ b/tests/nightly/estimator/test_sentiment_rnn.py
@@ -280,7 +280,3 @@ def test_estimator_gpu():
 
     assert acc.get()[1] > 0.70
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/nightly/test_kvstore.py b/tests/nightly/test_kvstore.py
old mode 100755
new mode 100644
diff --git a/tests/nightly/test_large_array.py b/tests/nightly/test_large_array.py
index 5fb0ff81da6b..0f0b373409b9 100644
--- a/tests/nightly/test_large_array.py
+++ b/tests/nightly/test_large_array.py
@@ -28,7 +28,6 @@
 from mxnet.test_utils import rand_ndarray, assert_almost_equal, rand_coord_2d, default_context, check_symbolic_forward, create_2d_tensor
 from mxnet import gluon, nd
 from common import with_seed, with_post_test_cleanup
-from nose.tools import with_setup
 import unittest
 
 # dimension constants
@@ -469,7 +468,7 @@ def check_col2im():
         assert res.shape[2] == 2
         assert res.shape[3] == 2
         assert res.shape[4] == 1
-        
+
     def check_embedding():
         data = nd.random_normal(shape=(LARGE_TENSOR_SHAPE, 1))
         weight = nd.random_normal(shape=(LARGE_TENSOR_SHAPE, 1))
@@ -480,7 +479,7 @@ def check_embedding():
 
         assert out.shape[0] == LARGE_TENSOR_SHAPE
         assert out.shape[1] == 1
-        
+
     def check_spatial_transformer():
         data = nd.random_normal(shape=(2, 2**29, 1, 6))
         loc = nd.random_normal(shape=(2, 6))
@@ -495,7 +494,7 @@ def check_spatial_transformer():
         assert res.shape[1] == 536870912
         assert res.shape[2] == 2
         assert res.shape[3] == 6
-        
+
     def check_ravel():
         data = nd.random_normal(shape=(2, LARGE_TENSOR_SHAPE))
         shape = (2, 10)
@@ -530,7 +529,7 @@ def check_multi_lars():
 
         # Trigger lazy evaluation of the output NDArray and ensure that it has been filled
         assert type(out[0, 0].asscalar()).__name__ == 'float32'
-        
+
     def check_rnn():
         data = nd.random_normal(shape=(RNN_LARGE_TENSOR, 4, 4))
         parameters_relu_tanh = nd.random_normal(shape=(7,))
@@ -547,10 +546,10 @@ def check_rnn():
 
         out_relu = nd.RNN(data=data, parameters=parameters_relu_tanh, state=state, mode=mode_relu,
                           state_size=state_size, num_layers=num_layers)
-        
+
         out_tanh = nd.RNN(data=data, parameters=parameters_relu_tanh, state=state, mode=mode_tanh,
                           state_size=state_size, num_layers=num_layers)
-        
+
         out_lstm = nd.RNN(data=data, parameters=parameters_lstm, state=state, mode=mode_lstm,
                           state_cell=state_cell, state_size=state_size, num_layers=num_layers)
 
@@ -1546,7 +1545,7 @@ def check_logical_xor(a, b):
     def create_input_for_rounding_ops():
         # Creates an vector with values (-LARGE_X/2 .... -2, -1, 0, 1, 2, .... , LARGE_X/2-1)
         # then divides each element by 2 i.e (-LARGE_X/4 .... -1, -0.5, 0, 0.5, 1, .... , LARGE_X/4-1)
-        # and finally broadcasts to 
+        # and finally broadcasts to
         inp = nd.arange(-LARGE_X//2, LARGE_X//2, dtype=np.float64).reshape(1, LARGE_X)
         inp = inp/2
         inp = nd.broadcast_to(inp, (SMALL_Y, LARGE_X))
@@ -1559,7 +1558,7 @@ def assert_correctness_of_rounding_ops(output, mid, expected_vals):
         for i in range(len(output_idx_to_inspect)):
             assert output[1][output_idx_to_inspect[i]] == expected_vals[i]
 
-    # TODO(access2rohit): merge similar tests in large vector and array into one file. 
+    # TODO(access2rohit): merge similar tests in large vector and array into one file.
     def check_rounding_ops():
         x = create_input_for_rounding_ops()
         def check_ceil():
@@ -1819,7 +1818,3 @@ def test_sparse_dot():
     assert out.asnumpy()[0][0] == 2
     assert out.shape == (2, 2)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/nightly/test_large_vector.py b/tests/nightly/test_large_vector.py
index bbad75627769..57fecaafc0ac 100644
--- a/tests/nightly/test_large_vector.py
+++ b/tests/nightly/test_large_vector.py
@@ -28,7 +28,6 @@
 from mxnet.test_utils import rand_ndarray, assert_almost_equal, rand_coord_2d, create_vector
 from mxnet import gluon, nd
 from tests.python.unittest.common import with_seed
-from nose.tools import with_setup
 import unittest
 
 # dimension constants
@@ -370,7 +369,7 @@ def check_slice_axis():
 
     def check_gather():
         arr = mx.nd.ones(LARGE_X)
-        # Passing dtype=np.int64 since randomly generated indices are 
+        # Passing dtype=np.int64 since randomly generated indices are
         # very large that exceeds int32 limits.
         idx = mx.nd.random.randint(0, LARGE_X, 10, dtype=np.int64)
         # Calls gather_nd internally
@@ -1063,7 +1062,3 @@ def check_minimum():
     check_maximum()
     check_minimum()
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/nightly/test_np_random.py b/tests/nightly/test_np_random.py
index 09ebdad90bd2..63ac4b9db60e 100644
--- a/tests/nightly/test_np_random.py
+++ b/tests/nightly/test_np_random.py
@@ -173,7 +173,3 @@ def test_np_laplace():
             generator_mx_np = lambda x: np.random.laplace(loc, scale, size=x, ctx=ctx, dtype=dtype).asnumpy()
             verify_generator(generator=generator_mx_np, buckets=buckets, probs=probs, nsamples=samples, nrepeat=trials)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/nightly/test_optimizer.py b/tests/nightly/test_optimizer.py
index c4e264b79b65..0a87368d991e 100644
--- a/tests/nightly/test_optimizer.py
+++ b/tests/nightly/test_optimizer.py
@@ -88,6 +88,3 @@ def test_lars():
     accuracy = acc.get()[1]
     assert accuracy > 0.98, "LeNet-5 training accuracy on MNIST was too low"
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_contrib_amp.py b/tests/python/gpu/test_contrib_amp.py
index 527f8534969c..398dfc394eff 100644
--- a/tests/python/gpu/test_contrib_amp.py
+++ b/tests/python/gpu/test_contrib_amp.py
@@ -24,14 +24,14 @@
 import collections
 import ctypes
 import mxnet.contrib.amp as amp
-from nose.tools import assert_raises
+import pytest
 from mxnet.test_utils import set_default_context, download_model, same_symbol_structure
 from mxnet.gluon.model_zoo.vision import get_model
 from mxnet.gluon import SymbolBlock, nn, rnn
 from mxnet.contrib.amp import amp
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import with_seed, teardown, assert_raises_cudnn_not_satisfied
+from common import with_seed, teardown_module, assert_raises_cudnn_not_satisfied
 sys.path.insert(0, os.path.join(curr_path, '../train'))
 from test_bucketing import train_model
 set_default_context(mx.gpu(0))
@@ -119,18 +119,18 @@ def check_amp_convert_symbol():
             "convert_symbol generating wrong computation graph"
 
         # convert_symbol called with incorrect inputs
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="float16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["elemwise_add"])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="float16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["Activation"],
                       conditional_fp32_ops=[('Activation', 'act_type', ['selu'])])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="float16", target_dtype_ops=["Activation"],
                       fp32_ops=["Activation"],
                       conditional_fp32_ops=[('Activation', 'act_type', ['selu'])])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="float16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["FullyConnected"])
 
@@ -484,7 +484,3 @@ def test_fp16_casting():
     out = mx.sym.split(concat_res, axis=1, num_outputs=2)
     final_res = amp.convert_symbol(out)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_deferred_compute_gpu.py b/tests/python/gpu/test_deferred_compute_gpu.py
index 7503c7ba102e..9802d2b57d24 100644
--- a/tests/python/gpu/test_deferred_compute_gpu.py
+++ b/tests/python/gpu/test_deferred_compute_gpu.py
@@ -24,10 +24,6 @@
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
 # We import all tests from ../unittest/test_deferred_compute.py
-# They will be detected by nose, as long as the current file has a different filename
+# They will be detected by test framework, as long as the current file has a different filename
 from test_deferred_compute import *
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_forward.py b/tests/python/gpu/test_forward.py
index 02b0256024d3..2ec5ee262e0c 100644
--- a/tests/python/gpu/test_forward.py
+++ b/tests/python/gpu/test_forward.py
@@ -22,7 +22,7 @@
 from mxnet.test_utils import *
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.gluon import utils
 import tarfile
 
diff --git a/tests/python/gpu/test_fusion.py b/tests/python/gpu/test_fusion.py
index c0a8bdbb0807..f3539c5d32f7 100644
--- a/tests/python/gpu/test_fusion.py
+++ b/tests/python/gpu/test_fusion.py
@@ -310,6 +310,3 @@ def test_fusion_reshape_executor():
     out = f.forward(is_train=False, data1=data, data2=data)
     assert out[0].sum().asscalar() == 150
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_gluon_contrib_gpu.py b/tests/python/gpu/test_gluon_contrib_gpu.py
index 348e9f77acc8..9b43c797a252 100644
--- a/tests/python/gpu/test_gluon_contrib_gpu.py
+++ b/tests/python/gpu/test_gluon_contrib_gpu.py
@@ -85,6 +85,3 @@ def test_ModulatedDeformableConvolution():
         y = net(x)
         y.backward()
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_gluon_gpu.py b/tests/python/gpu/test_gluon_gpu.py
index 7e90854d923f..4a398d1c6a4c 100644
--- a/tests/python/gpu/test_gluon_gpu.py
+++ b/tests/python/gpu/test_gluon_gpu.py
@@ -30,7 +30,7 @@
 
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown, assert_raises_cudnn_not_satisfied, run_in_spawned_process
+from common import setup_module, with_seed, teardown_module, assert_raises_cudnn_not_satisfied, run_in_spawned_process
 from test_gluon import *
 from test_loss import *
 from test_gluon_rnn import *
@@ -639,7 +639,3 @@ def test_gemms_true_fp16():
                         atol=atol, rtol=rtol)
     os.environ["MXNET_FC_TRUE_FP16"] = "0"
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_gluon_model_zoo_gpu.py b/tests/python/gpu/test_gluon_model_zoo_gpu.py
index 6f559db62808..ac50bb86e7ab 100644
--- a/tests/python/gpu/test_gluon_model_zoo_gpu.py
+++ b/tests/python/gpu/test_gluon_model_zoo_gpu.py
@@ -27,7 +27,7 @@
 import unittest
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 def eprint(*args, **kwargs):
     print(*args, file=sys.stderr, **kwargs)
@@ -180,6 +180,3 @@ def test_training():
                 gpu_param = gpu_params.get(k)
                 assert_almost_equal(cpu_param.data(), gpu_param.data(), rtol=1e-3, atol=1e-3)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_gluon_transforms.py b/tests/python/gpu/test_gluon_transforms.py
index e777f0bb90e2..23addbffc20f 100644
--- a/tests/python/gpu/test_gluon_transforms.py
+++ b/tests/python/gpu/test_gluon_transforms.py
@@ -27,7 +27,7 @@
 from mxnet.test_utils import almost_equal, same
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import assertRaises, setup_module, with_seed, teardown
+from common import assertRaises, setup_module, with_seed, teardown_module
 from test_gluon_data_vision import test_to_tensor, test_normalize, test_crop_resize
 
 set_default_context(mx.gpu(0))
diff --git a/tests/python/gpu/test_kvstore_gpu.py b/tests/python/gpu/test_kvstore_gpu.py
index a986f70d7525..f83220ad2e61 100644
--- a/tests/python/gpu/test_kvstore_gpu.py
+++ b/tests/python/gpu/test_kvstore_gpu.py
@@ -24,7 +24,7 @@
 from mxnet.test_utils import assert_almost_equal, default_context, EnvManager
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 shape = (4, 4)
 keys = [5, 7, 11]
@@ -40,12 +40,12 @@ def init_kv_with_str(stype='default', kv_type='local'):
     return kv
 
 # 1. Test seed 89411477 (module seed 1829754103) resulted in a py3-gpu CI runner core dump.
-# 2. Test seed 1155716252 (module seed 1032824746) resulted in py3-mkldnn-gpu have error 
+# 2. Test seed 1155716252 (module seed 1032824746) resulted in py3-mkldnn-gpu have error
 # src/operator/nn/mkldnn/mkldnn_base.cc:567: Check failed: similar
 # Both of them are not reproducible, so this test is back on random seeds.
 @with_seed()
 @unittest.skipIf(mx.context.num_gpus() < 2, "test_rsp_push_pull needs more than 1 GPU")
-@unittest.skip("Flaky test https://github.com/apache/incubator-mxnet/issues/14189") 
+@unittest.skip("Flaky test https://github.com/apache/incubator-mxnet/issues/14189")
 def test_rsp_push_pull():
     def check_rsp_push_pull(kv_type, sparse_pull, is_push_cpu=True):
         kv = init_kv_with_str('row_sparse', kv_type)
@@ -134,6 +134,3 @@ def test_rsp_push_pull_large_rowid():
     kv.row_sparse_pull('a', out=out, row_ids=mx.nd.arange(0, num_rows, dtype='int64'))
     assert(out.indices.shape[0] == num_rows)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_numpy_fallback.py b/tests/python/gpu/test_numpy_fallback.py
index 1499e1b15c57..71ca09c5ab2e 100644
--- a/tests/python/gpu/test_numpy_fallback.py
+++ b/tests/python/gpu/test_numpy_fallback.py
@@ -26,7 +26,6 @@
 import mxnet as mx
 import scipy.stats as ss
 import scipy.special as scipy_special
-from nose.tools import assert_raises
 from mxnet import np, npx
 from mxnet.base import MXNetError
 from mxnet.test_utils import assert_almost_equal, use_np, set_default_context
@@ -109,7 +108,3 @@ def empty_ret_func():
     # does not support functions with no return values
     assertRaises(ValueError, empty_ret_func)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_operator_gpu.py b/tests/python/gpu/test_operator_gpu.py
index a017b8ce59c2..82cefcd2f2b1 100644
--- a/tests/python/gpu/test_operator_gpu.py
+++ b/tests/python/gpu/test_operator_gpu.py
@@ -23,14 +23,14 @@
 import mxnet as mx
 import numpy as np
 import unittest
-from nose.tools import assert_raises
+import pytest
 from mxnet.test_utils import check_consistency, set_default_context, assert_almost_equal, assert_allclose
 from mxnet.base import MXNetError
 from mxnet import autograd
 
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown, assert_raises_cudnn_not_satisfied, assert_raises_cuda_not_satisfied
+from common import setup_module, with_seed, teardown_module, assert_raises_cudnn_not_satisfied, assert_raises_cuda_not_satisfied
 from common import run_in_spawned_process
 from test_operator import *
 from test_numpy_ndarray import *
@@ -1342,7 +1342,7 @@ def test_bilinear_resize_op():
     check_consistency(sym, ctx_list)
 
     sym = mx.sym.contrib.BilinearResize2D(data, height=10, width=5, align_corners=False)
-    check_consistency(sym, ctx_list)    
+    check_consistency(sym, ctx_list)
 
     sym = mx.sym.contrib.BilinearResize2D(data, None, scale_height=2, scale_width=0.5, mode='odd_scale', align_corners=True)
     check_consistency(sym, ctx_list)
@@ -2274,7 +2274,7 @@ def test_kernel_error_checking():
 
 def test_incorrect_gpu():
     # Try setting dev_id to a really big number
-    assert_raises(MXNetError, mx.nd.ones, (2,2), ctx=mx.gpu(100001))
+    pytest.raises(MXNetError, mx.nd.ones, (2,2), ctx=mx.gpu(100001))
 
 @with_seed()
 def test_batchnorm_backwards_notrain():
@@ -2526,7 +2526,7 @@ def run_math(op, shape, dtype="float32", check_value=True):
 def test_math():
     ops = ['log', 'erf', 'square']
     check_value= True
-    shape_lst = [[1000], [100,1000], [10,100,100], [10,100,100,100]] 
+    shape_lst = [[1000], [100,1000], [10,100,100], [10,100,100,100]]
     dtypes = ["float32", "float64"]
     for shape in shape_lst:
         for dtype in dtypes:
@@ -2548,6 +2548,3 @@ def test_arange_like_dtype():
         for v in out:
             assert v.dtype == t
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_predictor.py b/tests/python/gpu/test_predictor.py
index 4838a76c7cb1..592733a90174 100644
--- a/tests/python/gpu/test_predictor.py
+++ b/tests/python/gpu/test_predictor.py
@@ -31,7 +31,7 @@
 from mxnet.contrib.amp import amp
 from mxnet.base import NDArrayHandle, py_str
 sys.path.insert(0, os.path.join(curr_path, '../unittest'))
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 @with_seed()
 def test_predictor_with_dtype():
@@ -122,7 +122,3 @@ def test_predictor_amp():
                                                                          cast_optional_params=True)
     compare_module_cpredict(result_sym, result_arg_params, result_aux_params, monitor_callback=True)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/gpu/test_tvm_bridge.py b/tests/python/gpu/test_tvm_bridge.py
index 5c87536bdbae..7a4339c6ae94 100644
--- a/tests/python/gpu/test_tvm_bridge.py
+++ b/tests/python/gpu/test_tvm_bridge.py
@@ -61,7 +61,3 @@ def check(target, dtype):
                       "float32", "float64"]:
             check(tgt, dtype)
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/mkl/test_bf16_operator.py b/tests/python/mkl/test_bf16_operator.py
index b275c96bbb67..888b5d20b908 100644
--- a/tests/python/mkl/test_bf16_operator.py
+++ b/tests/python/mkl/test_bf16_operator.py
@@ -25,7 +25,6 @@
 import ctypes
 import itertools
 import mxnet.contrib.amp as amp
-from nose.tools import assert_raises
 from mxnet.test_utils import set_default_context, download_model, same_symbol_structure, assert_almost_equal_with_err, rand_shape_nd
 from mxnet.gluon.model_zoo.vision import get_model
 from mxnet.gluon import SymbolBlock, nn, rnn
@@ -55,7 +54,7 @@ def check_operator_accuracy(sym_fp32, sym_bf16, data_shape, num_input_data=1, bf
         the relative threshold
     atol: float
         the absolute threshold
-    etol: float 
+    etol: float
         The error rate threshold, allow a small amount of value not consistent between bf16 and fp32
     """
     if not isinstance(data_shape, tuple):
@@ -105,7 +104,7 @@ def check_operator_accuracy(sym_fp32, sym_bf16, data_shape, num_input_data=1, bf
             exe_bf16.arg_dict[arg_name][:] = arg_params_fp32[arg_name]
         else:
             exe_bf16.arg_dict[arg_name][:] = mx.nd.amp_cast(arg_params_fp32[arg_name], dtype=bfloat16)
-    
+
     for aux_name in aux_names:
         if bf16_use_fp32_params:
             exe_bf16.aux_dict[aux_name][:] = aux_params_fp32[aux_name]
@@ -169,7 +168,7 @@ def test_bf16_pooling():
     pool_conventions = ["full", "valid"]
     for new_params in itertools.product(data_shapes, pool_types, pool_conventions):
         pool_params.update({"pool_type": new_params[1], "pooling_convention": new_params[2]})
-        
+
         data_sym_fp32 = mx.sym.Variable(name='data')
         data_sym_bf16 = mx.sym.Variable(name='data', dtype=bfloat16)
         pool_fp32 = mx.sym.Pooling(data_sym_fp32, **pool_params)
@@ -230,7 +229,7 @@ def test_bf16_abs():
         data_sym_bf16 = mx.sym.Variable(name='data', dtype=bfloat16)
         sym_fp32 = mx.sym.abs(data_sym_fp32)
         sym_bf16 = mx.sym.abs(data_sym_bf16)
-        
+
         check_operator_accuracy(sym_fp32, sym_bf16, data_shape, bf16_use_fp32_params=True)
 
 @with_seed()
@@ -285,6 +284,3 @@ def test_bf16_fallback():
     conv_bf16 = mx.sym.Convolution(data_sym_bf16, **conv_params)
     check_operator_accuracy(sym_fp32=conv_fp32, sym_bf16=conv_bf16, data_shape=(3, 32, 28, 28, 4), bf16_use_fp32_params=False)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/mkl/test_contrib_amp.py b/tests/python/mkl/test_contrib_amp.py
index 5d5774099255..ec88851751ff 100644
--- a/tests/python/mkl/test_contrib_amp.py
+++ b/tests/python/mkl/test_contrib_amp.py
@@ -24,7 +24,7 @@
 import collections
 import ctypes
 import mxnet.contrib.amp as amp
-from nose.tools import assert_raises
+import pytest
 from mxnet.test_utils import set_default_context, download_model, same_symbol_structure, assert_almost_equal
 from mxnet.gluon.model_zoo.vision import get_model
 from mxnet.gluon import SymbolBlock, nn, rnn
@@ -117,18 +117,18 @@ def check_amp_convert_symbol():
             "convert_symbol generating wrong computation graph"
 
         # convert_symbol called with incorrect inputs
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="bfloat16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["elemwise_add"])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="bfloat16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["Activation"],
                       conditional_fp32_ops=[('Activation', 'act_type', ['selu'])])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="bfloat16", target_dtype_ops=["Activation"],
                       fp32_ops=["Activation"],
                       conditional_fp32_ops=[('Activation', 'act_type', ['selu'])])
-        assert_raises(AssertionError, amp.convert_symbol, res,
+        pytest.raises(AssertionError, amp.convert_symbol, res,
                       target_dtype="bfloat16", target_dtype_ops=["FullyConnected"],
                       fp32_ops=["FullyConnected"])
 
@@ -495,7 +495,3 @@ def test_bf16_casting():
     exe = final_res.simple_bind(ctx=mx.cpu(), data=(1, 2), data2=(1, 2))
     assert exe.arg_arrays[0].dtype == bfloat16
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/mkl/test_mkldnn.py b/tests/python/mkl/test_mkldnn.py
index b52bb03a80c8..59b860755750 100644
--- a/tests/python/mkl/test_mkldnn.py
+++ b/tests/python/mkl/test_mkldnn.py
@@ -642,6 +642,3 @@ def check_elemwise_add_training(stype):
     for stype in stypes:
         check_elemwise_add_training(stype)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/mkl/test_quantization_mkldnn.py b/tests/python/mkl/test_quantization_mkldnn.py
index 8ba2f2b01feb..f2432175720a 100644
--- a/tests/python/mkl/test_quantization_mkldnn.py
+++ b/tests/python/mkl/test_quantization_mkldnn.py
@@ -25,7 +25,7 @@
 from test_quantization import *
 
 if __name__ == '__main__':
-    import nose
-    nose.runmodule()
+    import pytest
+    pytest.main()
     del os.environ['ENABLE_MKLDNN_QUANTIZATION_TEST']
     del os.environ['MXNET_SUBGRAPH_BACKEND']
diff --git a/tests/python/mkl/test_subgraph.py b/tests/python/mkl/test_subgraph.py
index 65b73e438ea6..862fd9078c69 100644
--- a/tests/python/mkl/test_subgraph.py
+++ b/tests/python/mkl/test_subgraph.py
@@ -999,7 +999,3 @@ def test_quantized_fc_bias_overflow():
     helper_quantized_fc_bias_overflow(-1e-6, +1e-6, -1e-6, +1e-6)
     helper_quantized_fc_bias_overflow(0, 0, 0, 0)
 
-
-if __name__ == "__main__":
-  import nose
-  nose.runmodule()
diff --git a/tests/python/profiling/test_nvtx.py b/tests/python/profiling/test_nvtx.py
index 507b438e300d..a80e33ec03b0 100644
--- a/tests/python/profiling/test_nvtx.py
+++ b/tests/python/profiling/test_nvtx.py
@@ -46,7 +46,3 @@ def test_nvtx_ranges_present_in_profile():
     # Verify that we have some expected output from the engine.
     assert "Range \"WaitForVar\"" in profiler_output
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/quantization/test_quantization.py b/tests/python/quantization/test_quantization.py
index 8c6100d50765..d2179759ef8f 100644
--- a/tests/python/quantization/test_quantization.py
+++ b/tests/python/quantization/test_quantization.py
@@ -1253,7 +1253,3 @@ def get_threshold(nd):
         assert 'layer1' in th_dict
         assert_almost_equal(np.array([th_dict['layer1'][1]]), expected_threshold, rtol=1e-2, atol=1e-4)
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/quantization_gpu/test_quantization_gpu.py b/tests/python/quantization_gpu/test_quantization_gpu.py
index 4f2d70effd49..0f14fa1ac961 100644
--- a/tests/python/quantization_gpu/test_quantization_gpu.py
+++ b/tests/python/quantization_gpu/test_quantization_gpu.py
@@ -25,8 +25,3 @@
 from test_quantization import *
 
 set_default_context(mx.gpu(0))
-
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/tensorrt/lenet5_train.py b/tests/python/tensorrt/lenet5_train.py
old mode 100755
new mode 100644
diff --git a/tests/python/tensorrt/test_cvnets.py b/tests/python/tensorrt/test_cvnets.py
index 4b8eb48e926c..99312d76dc7a 100644
--- a/tests/python/tensorrt/test_cvnets.py
+++ b/tests/python/tensorrt/test_cvnets.py
@@ -167,8 +167,3 @@ def test_tensorrt_on_cifar_resnets(batch_size=32, tolerance=0.1, num_workers=1):
     finally:
         mx.contrib.tensorrt.set_use_fp16(original_use_fp16)
 
-
-if __name__ == '__main__':
-    import nose
-
-    nose.runmodule()
diff --git a/tests/python/tensorrt/test_ops.py b/tests/python/tensorrt/test_ops.py
index af1c453111d9..dfbbb8e8883c 100644
--- a/tests/python/tensorrt/test_ops.py
+++ b/tests/python/tensorrt/test_ops.py
@@ -512,6 +512,3 @@ def test_dropout():
             sym = mx.sym.Dropout(data, p=0.7, mode=mode, axes=(0,))
             check_unsupported_single_sym(sym)
 
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/tensorrt/test_resnet18.py b/tests/python/tensorrt/test_resnet18.py
index 9fd99abb121b..e146423e257d 100644
--- a/tests/python/tensorrt/test_resnet18.py
+++ b/tests/python/tensorrt/test_resnet18.py
@@ -69,6 +69,3 @@ def test_tensorrt_resnet18_feature_vect():
     finally:
         mx.contrib.tensorrt.set_use_fp16(original_precision_value)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/tensorrt/test_tensorrt_lenet5.py b/tests/python/tensorrt/test_tensorrt_lenet5.py
index 78f41ca53909..a34b634121d2 100644
--- a/tests/python/tensorrt/test_tensorrt_lenet5.py
+++ b/tests/python/tensorrt/test_tensorrt_lenet5.py
@@ -116,6 +116,3 @@ def test_tensorrt_inference():
         """Absolute diff. between MXNet & TensorRT accuracy (%f) exceeds threshold (%f):
            MXNet = %f, TensorRT = %f""" % (absolute_accuracy_diff, epsilon, mx_pct, trt_pct)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/common.py b/tests/python/unittest/common.py
index ab2d191f1360..fda43673003b 100644
--- a/tests/python/unittest/common.py
+++ b/tests/python/unittest/common.py
@@ -29,9 +29,33 @@
 
 import models
 from contextlib import contextmanager
-from nose.tools import make_decorator, assert_raises
+import pytest
 import tempfile
 
+def make_decorator(func):
+    """
+    Wraps a test decorator so as to properly replicate metadata
+    of the decorated function, including test tool additional stuff
+    (namely, setup and teardown).
+    """
+    def decorate(newfunc):
+        if hasattr(func, 'compat_func_name'):
+            name = func.compat_func_name
+        else:
+            name = func.__name__
+        newfunc.__dict__ = func.__dict__
+        newfunc.__doc__ = func.__doc__
+        newfunc.__module__ = func.__module__
+        if not hasattr(newfunc, 'compat_co_firstlineno'):
+            newfunc.compat_co_firstlineno = func.__code__.co_firstlineno
+        try:
+            newfunc.__name__ = name
+        except TypeError:
+            # can't set func name in 2.3
+            newfunc.compat_func_name = name
+        return newfunc
+    return decorate
+
 def assertRaises(expected_exception, func, *args, **kwargs):
     try:
         func(*args, **kwargs)
@@ -43,7 +67,7 @@ def assertRaises(expected_exception, func, *args, **kwargs):
 
 
 def default_logger():
-    """A logger used to output seed information to nosetests logs."""
+    """A logger used to output seed information to logs."""
     logger = logging.getLogger(__name__)
     # getLogger() lookups will return the same logger, but only add the handler once.
     if not len(logger.handlers):
@@ -109,7 +133,7 @@ def less_than(version_left, version_right):
         left = version_left.split(".")
         right = version_right.split(".")
 
-        # 0 pad shortest version - e.g. 
+        # 0 pad shortest version - e.g.
         # less_than("9.1", "9.1.9") == less_than("9.1.0", "9.1.9")
         longest = max(len(left), len(right))
         left.extend([0] * (longest - len(left)))
@@ -131,7 +155,7 @@ def test_new(*args, **kwargs):
             if not cuxx_test_disabled or mx.context.current_context().device_type == 'cpu':
                 orig_test(*args, **kwargs)
             else:
-                assert_raises((MXNetError, RuntimeError), orig_test, *args, **kwargs)
+                pytest.raises((MXNetError, RuntimeError), orig_test, *args, **kwargs)
         return test_new
     return test_helper
 
@@ -154,7 +178,7 @@ def assert_raises_cuda_not_satisfied(min_version):
 
 def with_seed(seed=None):
     """
-    A decorator for nosetests test functions that manages rng seeds.
+    A decorator for test functions that manages rng seeds.
 
     Parameters
     ----------
@@ -181,10 +205,10 @@ def test_not_ok_with_random_data():
     can then set the environment variable MXNET_TEST_SEED to
     the value reported, then rerun the test with:
 
-        nosetests --verbose -s <test_module_name.py>:<failing_test>
+        pytest --verbose --capture=no <test_module_name.py>::<failing_test>
 
     To run a test repeatedly, set MXNET_TEST_COUNT=<NNN> in the environment.
-    To see the seeds of even the passing tests, add '--logging-level=DEBUG' to nosetests.
+    To see the seeds of even the passing tests, add '--log-level=DEBUG' to pytest.
     """
     def test_helper(orig_test):
         @make_decorator(orig_test)
@@ -206,7 +230,7 @@ def test_new(*args, **kwargs):
                 mx.random.seed(this_test_seed)
                 random.seed(this_test_seed)
                 logger = default_logger()
-                # 'nosetests --logging-level=DEBUG' shows this msg even with an ensuing core dump.
+                # 'pytest --logging-level=DEBUG' shows this msg even with an ensuing core dump.
                 test_count_msg = '{} of {}: '.format(i+1,test_count) if test_count > 1 else ''
                 test_msg = ('{}Setting test np/mx/python random seeds, use MXNET_TEST_SEED={}'
                             ' to reproduce.').format(test_count_msg, this_test_seed)
@@ -226,12 +250,12 @@ def test_new(*args, **kwargs):
 
 def setup_module():
     """
-    A function with a 'magic name' executed automatically before each nosetests module
+    A function with a 'magic name' executed automatically before each pytest module
     (file of tests) that helps reproduce a test segfault by setting and outputting the rng seeds.
 
     The segfault-debug procedure on a module called test_module.py is:
 
-    1. run "nosetests --verbose test_module.py".  A seg-faulting output might be:
+    1. run "pytest --verbose test_module.py".  A seg-faulting output might be:
 
        [INFO] np, mx and python random seeds = 4018804151
        test_module.test1 ... ok
@@ -239,7 +263,7 @@ def setup_module():
 
     2. Copy the module-starting seed into the next command, then run:
 
-       MXNET_MODULE_SEED=4018804151 nosetests --logging-level=DEBUG --verbose test_module.py
+       MXNET_MODULE_SEED=4018804151 pytest --logging-level=DEBUG --verbose test_module.py
 
        Output might be:
 
@@ -251,7 +275,7 @@ def setup_module():
        Illegal instruction (core dumped)
 
     3. Copy the segfaulting-test seed into the command:
-       MXNET_TEST_SEED=1435005594 nosetests --logging-level=DEBUG --verbose test_module.py:test2
+       MXNET_TEST_SEED=1435005594 pytest --logging-level=DEBUG --verbose test_module.py:test2
        Output might be:
 
        [INFO] np, mx and python random seeds = 2481884723
@@ -301,9 +325,9 @@ def __enter__(self):
         def __exit__(self, exc_type, exc_value, traceback):
             shutil.rmtree(self._dirname)
 
-def teardown():
+def teardown_module():
     """
-    A function with a 'magic name' executed automatically after each nosetests test module.
+    A function with a 'magic name' executed automatically after each pytest test module.
 
     It waits for all operations in one file to finish before carrying on the next.
     """
@@ -373,4 +397,4 @@ def run_in_spawned_process(func, env, *args):
         finally:
             os.environ.clear()
             os.environ.update(orig_environ)
-    return True
\ No newline at end of file
+    return True
diff --git a/tests/python-pytest/onnx/README.md b/tests/python/unittest/onnx/README.md
similarity index 100%
rename from tests/python-pytest/onnx/README.md
rename to tests/python/unittest/onnx/README.md
diff --git a/tests/python-pytest/onnx/backend.py b/tests/python/unittest/onnx/backend.py
similarity index 100%
rename from tests/python-pytest/onnx/backend.py
rename to tests/python/unittest/onnx/backend.py
diff --git a/tests/python-pytest/onnx/backend_rep.py b/tests/python/unittest/onnx/backend_rep.py
similarity index 100%
rename from tests/python-pytest/onnx/backend_rep.py
rename to tests/python/unittest/onnx/backend_rep.py
diff --git a/tests/python-pytest/onnx/backend_test.py b/tests/python/unittest/onnx/backend_test.py
old mode 100755
new mode 100644
similarity index 100%
rename from tests/python-pytest/onnx/backend_test.py
rename to tests/python/unittest/onnx/backend_test.py
diff --git a/tests/python-pytest/onnx/mxnet_export_test.py b/tests/python/unittest/onnx/mxnet_export_test.py
similarity index 100%
rename from tests/python-pytest/onnx/mxnet_export_test.py
rename to tests/python/unittest/onnx/mxnet_export_test.py
diff --git a/tests/python-pytest/onnx/test_cases.py b/tests/python/unittest/onnx/test_cases.py
similarity index 100%
rename from tests/python-pytest/onnx/test_cases.py
rename to tests/python/unittest/onnx/test_cases.py
diff --git a/tests/python-pytest/onnx/test_models.py b/tests/python/unittest/onnx/test_models.py
similarity index 100%
rename from tests/python-pytest/onnx/test_models.py
rename to tests/python/unittest/onnx/test_models.py
diff --git a/tests/python-pytest/onnx/test_node.py b/tests/python/unittest/onnx/test_node.py
similarity index 100%
rename from tests/python-pytest/onnx/test_node.py
rename to tests/python/unittest/onnx/test_node.py
diff --git a/tests/python/unittest/test_autograd.py b/tests/python/unittest/test_autograd.py
index 61955f034a71..69b61b47d4c9 100644
--- a/tests/python/unittest/test_autograd.py
+++ b/tests/python/unittest/test_autograd.py
@@ -20,7 +20,7 @@
 from mxnet.ndarray import zeros_like
 from mxnet.autograd import *
 from mxnet.test_utils import *
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.test_utils import EnvManager
 
 
@@ -467,7 +467,3 @@ def test_gradient():
     dx.backward()
     assert abs(x.grad.asscalar() - 2.71828175) < 1e-7
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_base.py b/tests/python/unittest/test_base.py
index 3189729e1d10..07d429589ba2 100644
--- a/tests/python/unittest/test_base.py
+++ b/tests/python/unittest/test_base.py
@@ -17,7 +17,6 @@
 
 import mxnet as mx
 from mxnet.base import data_dir
-from nose.tools import *
 import os
 import unittest
 import logging
@@ -47,4 +46,3 @@ def test_data_dir(self,):
         del os.environ['MXNET_HOME']
         self.assertEqual(data_dir(), prev_data_dir)
 
-
diff --git a/tests/python/unittest/test_contrib_autograd.py b/tests/python/unittest/test_contrib_autograd.py
index 1c878e322e7c..c376eb7794f8 100644
--- a/tests/python/unittest/test_contrib_autograd.py
+++ b/tests/python/unittest/test_contrib_autograd.py
@@ -18,7 +18,7 @@
 import mxnet.ndarray as nd
 from mxnet.contrib.autograd import *
 from mxnet.test_utils import *
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 def autograd_assert(*args, **kwargs):
     func   = kwargs["func"]
@@ -190,7 +190,3 @@ def test_retain_grad():
     raise AssertionError(
         "differentiating the same graph twice without retain_graph should fail")
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_control_flow.py b/tests/python/unittest/test_contrib_control_flow.py
index a93c109d11df..6db1003fdd20 100644
--- a/tests/python/unittest/test_contrib_control_flow.py
+++ b/tests/python/unittest/test_contrib_control_flow.py
@@ -2168,6 +2168,3 @@ def test_foreach_with_unkown_dim():
     _, output_shape, _ = outs.infer_shape_partial()
     assert_allclose((0, 3, 32, 32), output_shape[0])
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_hawkesll.py b/tests/python/unittest/test_contrib_hawkesll.py
index a4b1d9de605f..8f02737e2d12 100644
--- a/tests/python/unittest/test_contrib_hawkesll.py
+++ b/tests/python/unittest/test_contrib_hawkesll.py
@@ -154,8 +154,3 @@ def test_hawkesll_backward_single_mark():
 
     assert np.allclose(beta.grad.asnumpy().sum(), -0.05371582)
 
-
-if __name__ == "__main__":
-    import nose
-
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_operator.py b/tests/python/unittest/test_contrib_operator.py
index 476dfac24a61..717ce7f20f95 100644
--- a/tests/python/unittest/test_contrib_operator.py
+++ b/tests/python/unittest/test_contrib_operator.py
@@ -444,7 +444,3 @@ def test_modulated_deformable_convolution():
                         else:
                             rtol, atol = 0.05, 1e-3
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_optimizer.py b/tests/python/unittest/test_contrib_optimizer.py
index 5f7c51f257b3..aae767202454 100644
--- a/tests/python/unittest/test_contrib_optimizer.py
+++ b/tests/python/unittest/test_contrib_optimizer.py
@@ -196,6 +196,3 @@ def run_adamw_test(nElem=1, aggregate=False):
     for nElem in range(6):
         run_adamw_test(nElem+1)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_stes_op.py b/tests/python/unittest/test_contrib_stes_op.py
index 5864ec9db5b1..163ab4a7c9f8 100644
--- a/tests/python/unittest/test_contrib_stes_op.py
+++ b/tests/python/unittest/test_contrib_stes_op.py
@@ -132,6 +132,3 @@ def test_contrib_sign_ste():
     check_ste(net_type_str="SignSTENET", w_init=w_init, hybridize=True, in_data=in_data)
     check_ste(net_type_str="SignSTENET", w_init=w_init, hybridize=False, in_data=in_data)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
\ No newline at end of file
diff --git a/tests/python/unittest/test_contrib_svrg_module.py b/tests/python/unittest/test_contrib_svrg_module.py
index 79407d15fd7f..f135255753ce 100644
--- a/tests/python/unittest/test_contrib_svrg_module.py
+++ b/tests/python/unittest/test_contrib_svrg_module.py
@@ -307,7 +307,3 @@ def test_fit():
     estimated_mse = 1e-5
     assert metric.get()[1] < estimated_mse
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_svrg_optimizer.py b/tests/python/unittest/test_contrib_svrg_optimizer.py
index f7d90d12872f..cb6fdcf1218a 100644
--- a/tests/python/unittest/test_contrib_svrg_optimizer.py
+++ b/tests/python/unittest/test_contrib_svrg_optimizer.py
@@ -95,7 +95,3 @@ def test_kvstore_init_aux_keys():
     # updated with AssignmentOptimizer
     assert same(param_weight_full_init.asnumpy(), np.array([2, 2, 2]))
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_contrib_text.py b/tests/python/unittest/test_contrib_text.py
index 4072cc84f684..0778e5ace1f7 100644
--- a/tests/python/unittest/test_contrib_text.py
+++ b/tests/python/unittest/test_contrib_text.py
@@ -792,7 +792,3 @@ def test_get_and_pretrain_file_names():
 
     assertRaises(KeyError, text.embedding.get_pretrained_file_names, 'unknown$$')
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_deferred_compute.py b/tests/python/unittest/test_deferred_compute.py
index cebb6908768f..6d250d5fa468 100644
--- a/tests/python/unittest/test_deferred_compute.py
+++ b/tests/python/unittest/test_deferred_compute.py
@@ -19,11 +19,11 @@
 import operator
 
 import numpy as np
-from nose.tools import raises
 
 import mxnet as mx
 import mxnet._deferred_compute as dc
 from mxnet.base import MXNetError
+import pytest
 
 
 def _all_same(arrays1, arrays2, message=''):
@@ -213,7 +213,7 @@ def f(a, *, nd):
     _all_assert_dc(_dc_simple_setup, f)
 
 
-@raises(MXNetError)  # Should raise NotImplementedError https://github.com/apache/incubator-mxnet/issues/17522
+@pytest.mark.xfail(raises=MXNetError)  # Should raise NotImplementedError https://github.com/apache/incubator-mxnet/issues/17522
 def test_dc_inplace():
     def f(a, *, nd):
         a[:5] = 0
@@ -245,7 +245,7 @@ def test_dc_get_symbol_called_twice():
     assert sym2.list_inputs() == ['a']
 
 
-@raises(MXNetError)  # Should raise ValueError https://github.com/apache/incubator-mxnet/issues/17522
+@pytest.mark.xfail(raises=MXNetError)  # Should raise ValueError https://github.com/apache/incubator-mxnet/issues/17522
 def test_dc_set_variable_called_twice():
     a = mx.np.arange(10)
     dc.set_variable(a, mx.sym.var('a'))
@@ -342,7 +342,7 @@ def f(a, idx, *, nd):
         _assert_dc(setup, f, mode=mode)
 
 
-@raises(TypeError)  # Advanced indexing
+@pytest.mark.xfail(raises=TypeError)  # Advanced indexing
 def test_dc_list_indexing():
     def f(a, *, nd):
         assert nd is mx.np
@@ -352,7 +352,7 @@ def f(a, *, nd):
         _assert_dc(_dc_simple_setup, f, mode=mode)
 
 
-@raises(TypeError)  # Advanced indexing
+@pytest.mark.xfail(raises=TypeError)  # Advanced indexing
 def test_dc_numpy_indexing():
     def f(a, *, nd):
         assert nd is mx.np
@@ -430,7 +430,7 @@ def forward(self, x):
         _assert_dc_gluon(_dc_gluon_simple_setup, net, numpy=True)
 
 
-@raises(RuntimeError)
+@pytest.mark.xfail(raises=RuntimeError)
 def test_dc_hybridblock_deferred_init_no_infer_shape():
     class MyBlock(mx.gluon.HybridBlock):
         def __init__(self, *, prefix=None, params=None):
@@ -496,7 +496,7 @@ def setup(*, nd):
         _assert_dc_gluon(setup, net, numpy=True)
 
 
-@raises(RuntimeError)
+@pytest.mark.xfail(raises=RuntimeError)
 def test_dc_hybridblock_symbolblock():
     model = mx.gluon.nn.HybridSequential()
     model.add(mx.gluon.nn.Dense(128, activation='tanh'))
@@ -530,7 +530,3 @@ def forward(self, x):
 
     _all_same([out], [out_hybrid])
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_dgl_graph.py b/tests/python/unittest/test_dgl_graph.py
index 805adc2dac6f..89533fb119aa 100644
--- a/tests/python/unittest/test_dgl_graph.py
+++ b/tests/python/unittest/test_dgl_graph.py
@@ -240,6 +240,3 @@ def test_adjacency():
     assert_array_equal(adj.indices, g.indices)
     assert_array_equal(adj.data, mx.nd.ones(shape=g.indices.shape))
 
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_dlpack.py b/tests/python/unittest/test_dlpack.py
index fb64f8d58831..46bdde7d0bcd 100644
--- a/tests/python/unittest/test_dlpack.py
+++ b/tests/python/unittest/test_dlpack.py
@@ -43,6 +43,3 @@ def from_dlpack_old(dlpack):
     z = from_dlpack_old(y)
     assert_almost_equal(x.asnumpy(), z.asnumpy(), rtol=1e-5, atol=1e-5)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_dynamic_shape.py b/tests/python/unittest/test_dynamic_shape.py
index 1b043c73256d..b61fbeef6b74 100644
--- a/tests/python/unittest/test_dynamic_shape.py
+++ b/tests/python/unittest/test_dynamic_shape.py
@@ -48,7 +48,3 @@ def hybrid_forward(self, F, data, index):
     assert_almost_equal(result.asnumpy(), result_nd)
     assert_almost_equal(data.grad.asnumpy(), data_grad_nd)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_engine.py b/tests/python/unittest/test_engine.py
index 61d94ddbf4ec..fafc6758d892 100644
--- a/tests/python/unittest/test_engine.py
+++ b/tests/python/unittest/test_engine.py
@@ -15,7 +15,6 @@
 # specific language governing permissions and limitations
 # under the License.
 
-import nose
 import mxnet as mx
 import os
 import unittest
@@ -70,8 +69,3 @@ def test_engine_openmp_after_fork():
             print("Child omp max threads: {}".format(omp_max_threads))
             assert omp_max_threads == 1
 
-
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_engine_import.py b/tests/python/unittest/test_engine_import.py
index 303f3ceb1dee..7675cf836999 100644
--- a/tests/python/unittest/test_engine_import.py
+++ b/tests/python/unittest/test_engine_import.py
@@ -25,7 +25,7 @@
 
 def test_engine_import():
     import mxnet
-        
+
     engine_types = ['', 'NaiveEngine', 'ThreadedEngine', 'ThreadedEnginePerDevice']
 
     for type in engine_types:
@@ -35,7 +35,3 @@ def test_engine_import():
             os.environ.pop('MXNET_ENGINE_TYPE', None)
         reload(mxnet)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_exc_handling.py b/tests/python/unittest/test_exc_handling.py
index e3c333705260..8657bec1398c 100644
--- a/tests/python/unittest/test_exc_handling.py
+++ b/tests/python/unittest/test_exc_handling.py
@@ -18,11 +18,11 @@
 import mxnet as mx
 import numpy as np
 from mxnet import gluon
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.gluon import nn
 from mxnet.base import MXNetError
 from mxnet.test_utils import assert_exception, default_context, set_default_context, use_np
-from nose.tools import assert_raises
+import pytest
 
 
 @with_seed()
@@ -35,7 +35,7 @@ def imperative(exec_numpy=True):
             c.asnumpy()
 
     imperative(exec_numpy=False)
-    assert_raises(MXNetError, imperative, exec_numpy=True)
+    pytest.raises(MXNetError, imperative, exec_numpy=True)
 
 @with_seed()
 def test_exc_symbolic():
@@ -69,11 +69,11 @@ def symbolic(exec_backward=True, waitall=True):
             else:
                 outputs[0].asnumpy()
 
-    assert_raises(MXNetError, symbolic, exec_backward=False)
-    assert_raises(MXNetError, symbolic, exec_backward=True)
+    pytest.raises(MXNetError, symbolic, exec_backward=False)
+    pytest.raises(MXNetError, symbolic, exec_backward=True)
 
-    assert_raises(MXNetError, symbolic, exec_backward=False, waitall=True)
-    assert_raises(MXNetError, symbolic, exec_backward=True, waitall=True)
+    pytest.raises(MXNetError, symbolic, exec_backward=False, waitall=True)
+    pytest.raises(MXNetError, symbolic, exec_backward=True, waitall=True)
 
 @with_seed()
 def test_exc_gluon():
@@ -93,9 +93,9 @@ def gluon(exec_wait=True, waitall=False):
             z.wait_to_read()
 
     gluon(exec_wait=False)
-    assert_raises(MXNetError, gluon, exec_wait=True)
+    pytest.raises(MXNetError, gluon, exec_wait=True)
 
-    assert_raises(MXNetError, gluon, waitall=True)
+    pytest.raises(MXNetError, gluon, waitall=True)
 
 @with_seed()
 def test_exc_multiple_waits():
@@ -152,8 +152,8 @@ def mutable_var_check(waitall=False):
             mx.nd.waitall()
         else:
             a.asnumpy()
-    assert_raises(MXNetError, mutable_var_check, waitall=False)
-    assert_raises(MXNetError, mutable_var_check, waitall=True)
+    pytest.raises(MXNetError, mutable_var_check, waitall=False)
+    pytest.raises(MXNetError, mutable_var_check, waitall=True)
 
 @with_seed()
 def test_multiple_waitalls():
@@ -189,16 +189,16 @@ def check_resize():
         img = mx.nd.ones((1200, 1600, 3))
         img = mx.image.imresize(img, 320, 320, interp=-1)
         img.asnumpy()
-    assert_raises(MXNetError, check_resize)
+    pytest.raises(MXNetError, check_resize)
 
 
 @with_seed()
 def test_np_reshape_exception():
     a = mx.np.ones((10, 10))
     a.reshape((-1,)).asnumpy()  # Check no-raise
-    assert_raises(MXNetError, lambda: a.reshape((1,)))
-    assert_raises(MXNetError, lambda: mx.np.reshape(a, (1,)))
-    assert_raises(MXNetError, lambda: mx.np.reshape(a, (-1, 3)))
+    pytest.raises(MXNetError, lambda: a.reshape((1,)))
+    pytest.raises(MXNetError, lambda: mx.np.reshape(a, (1,)))
+    pytest.raises(MXNetError, lambda: mx.np.reshape(a, (-1, 3)))
 
 
 @with_seed()
@@ -208,10 +208,6 @@ def test_np_random_incorrect_named_arguments():
     for op_name in random_ops:
         op = getattr(mx.np.random, op_name, None)
         assert op is not None
-        assert_raises(TypeError, op, shape=())
-        assert_raises(TypeError, op, shape=None)
+        pytest.raises(TypeError, op, shape=())
+        pytest.raises(TypeError, op, shape=None)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_executor.py b/tests/python/unittest/test_executor.py
index 2bc696fd4e43..300e4b2590c8 100644
--- a/tests/python/unittest/test_executor.py
+++ b/tests/python/unittest/test_executor.py
@@ -17,7 +17,7 @@
 
 import numpy as np
 import mxnet as mx
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.test_utils import assert_almost_equal
 
 
@@ -164,7 +164,3 @@ def test_reshape():
     # weight ndarray is shared between exe and new_exe
     assert np.all(new_exe.arg_arrays[1].asnumpy() == 1)
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon.py b/tests/python/unittest/test_gluon.py
index 453c31ebc5de..cca1cfdce381 100644
--- a/tests/python/unittest/test_gluon.py
+++ b/tests/python/unittest/test_gluon.py
@@ -27,11 +27,11 @@
 from mxnet.ndarray.ndarray import _STORAGE_TYPE_STR_TO_ID
 from mxnet.test_utils import use_np
 import mxnet.numpy as _mx_np
-from common import (setup_module, with_seed, assertRaises, teardown,
+from common import (setup_module, with_seed, assertRaises, teardown_module,
                     assert_raises_cudnn_not_satisfied)
 import numpy as np
 from numpy.testing import assert_array_equal
-from nose.tools import raises, assert_raises
+import pytest
 from copy import deepcopy
 import warnings
 import json
@@ -54,12 +54,12 @@ def test_parameter():
     assert p.list_ctx() == [mx.cpu(1), mx.cpu(2)]
 
 @with_seed()
-@raises(AssertionError)
+@pytest.mark.xfail(raises=AssertionError)
 def test_invalid_parameter_stype():
     p = gluon.Parameter('weight', shape=(10, 10), stype='invalid')
 
 @with_seed()
-@raises(AssertionError)
+@pytest.mark.xfail(raises=AssertionError)
 def test_invalid_parameter_grad_stype():
     p = gluon.Parameter('weight', shape=(10, 10), grad_stype='invalid')
 
@@ -424,7 +424,7 @@ def hybrid_forward(self, F, x):
     assert np.dtype(prediction.dtype) == np.dtype(np.float32)
 
 @with_seed()
-@raises(AssertionError)
+@pytest.mark.xfail(raises=AssertionError)
 def test_sparse_symbol_block():
     data = mx.sym.var('data')
     weight = mx.sym.var('weight', stype='row_sparse')
@@ -434,7 +434,7 @@ def test_sparse_symbol_block():
     net = gluon.SymbolBlock(out, data)
 
 @with_seed()
-@raises(RuntimeError)
+@pytest.mark.xfail(raises=RuntimeError)
 def test_sparse_hybrid_block():
     params = gluon.ParameterDict('net_')
     params.get('weight', shape=(5,5), stype='row_sparse', dtype='float32')
@@ -501,17 +501,17 @@ def hybrid_forward(self, F, a, b):
         foo = FooNested()
         if do_hybridize:
             foo.hybridize()
-        assert_raises(ValueError, foo, None, None)
+        pytest.raises(ValueError, foo, None, None)
 
     # Make sure the ValueError is correctly raised
     foo = FooNested()
     foo.hybridize()
     foo(None, mx.nd.ones((10,)))  # Pass for the first time to initialize the cached op
-    assert_raises(ValueError, lambda: foo(mx.nd.ones((10,)), mx.nd.ones((10,))))
+    pytest.raises(ValueError, lambda: foo(mx.nd.ones((10,)), mx.nd.ones((10,))))
     foo = FooNested()
-    assert_raises(ValueError, lambda: foo(mx.nd.ones((10,)), mx.sym.var('a')))
+    pytest.raises(ValueError, lambda: foo(mx.nd.ones((10,)), mx.sym.var('a')))
     foo = FooNested()
-    assert_raises(ValueError, lambda: foo(mx.sym.var('a'), mx.nd.ones((10,))))
+    pytest.raises(ValueError, lambda: foo(mx.sym.var('a'), mx.nd.ones((10,))))
 
     # Test the case of the default values
     foo1 = FooDefault()
@@ -529,7 +529,7 @@ def hybrid_forward(self, F, a, b):
     out1 = foo1(mx.nd.ones((10,)), None)
     out2 = foo1(mx.nd.ones((10,)))
     assert_almost_equal(out1.asnumpy(), out2.asnumpy())
-    assert_raises(ValueError, lambda: foo1(mx.nd.ones((10,)), mx.nd.ones((10,))))
+    pytest.raises(ValueError, lambda: foo1(mx.nd.ones((10,)), mx.nd.ones((10,))))
 
 
 @with_seed()
@@ -567,13 +567,13 @@ def forward(self, a, b):
     # 4. Allow mixing of cpu_pinned and cpu
     foo_hybrid = FooHybrid()
     foo_hybrid.hybridize()
-    assert_raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,)), 1))
+    pytest.raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,)), 1))
     foo_hybrid = FooHybrid()
     foo_hybrid.hybridize()
-    assert_raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,)), mx.sym.var('a')))
+    pytest.raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,)), mx.sym.var('a')))
     foo_hybrid = FooHybrid()
     foo_hybrid.hybridize()
-    assert_raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,), ctx=mx.cpu(1)),
+    pytest.raises(ValueError, lambda: foo_hybrid(mx.nd.ones((10,), ctx=mx.cpu(1)),
                                                  mx.nd.ones((10,), ctx=mx.cpu(2))))
 
 
@@ -900,7 +900,7 @@ def test_layernorm():
         layer.initialize()
         if hybridize:
             layer.hybridize()
-        assert_raises(MXNetError, lambda: layer(mx.nd.ones((2, 11))))
+        pytest.raises(MXNetError, lambda: layer(mx.nd.ones((2, 11))))
 
 @with_seed()
 def test_groupnorm():
@@ -1023,7 +1023,7 @@ def test_block_attr_hidden():
     b.a = 1
 
 
-@raises(TypeError)
+@pytest.mark.xfail(raises=TypeError)
 @with_seed()
 def test_block_attr_block():
     b = gluon.Block()
@@ -1033,7 +1033,7 @@ def test_block_attr_block():
     b.b = (2,)
 
 
-@raises(TypeError)
+@pytest.mark.xfail(raises=TypeError)
 @with_seed()
 def test_block_attr_param():
     b = gluon.Block()
@@ -1885,7 +1885,7 @@ def test_summary():
     net3.summary(mx.nd.ones((80, 32, 5)), begin_state)
 
     net.hybridize()
-    assert_raises(AssertionError, net.summary, mx.nd.ones((32, 3, 224, 224)))
+    pytest.raises(AssertionError, net.summary, mx.nd.ones((32, 3, 224, 224)))
 
 
 @with_seed()
@@ -3214,6 +3214,3 @@ def hybrid_forward(self, F, x):
 
     mx.test_utils.assert_almost_equal(grad1, grad2)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_batch_processor.py b/tests/python/unittest/test_gluon_batch_processor.py
index 8604713fc129..952ed1c4a0da 100644
--- a/tests/python/unittest/test_gluon_batch_processor.py
+++ b/tests/python/unittest/test_gluon_batch_processor.py
@@ -27,7 +27,7 @@
 from mxnet.gluon.contrib.estimator import *
 from mxnet.gluon.contrib.estimator.event_handler import *
 from mxnet.gluon.contrib.estimator.batch_processor import BatchProcessor
-from nose.tools import assert_raises
+import pytest
 
 def _get_test_network():
     net = nn.Sequential()
@@ -66,12 +66,12 @@ def test_batch_processor_fit():
     est.fit(train_data=dataloader,
             epochs=num_epochs)
 
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=dataiter,
                 epochs=num_epochs)
 
     # Input NDArray
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=[mx.nd.ones(shape=(10, 3))],
                 epochs=num_epochs)
 
@@ -105,12 +105,12 @@ def test_batch_processor_validation():
     val_metrics = est.val_metrics
     validation_handler = ValidationHandler(val_data=dataloader, eval_fn=est.evaluate)
 
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=dataiter,
                 val_data=dataiter,
                 epochs=num_epochs)
     # Input NDArray
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=[mx.nd.ones(shape=(10, 3))],
                 val_data=[mx.nd.ones(shape=(10, 3))],
                 epochs=num_epochs)
diff --git a/tests/python/unittest/test_gluon_contrib.py b/tests/python/unittest/test_gluon_contrib.py
index 0ed0d4e8a545..b77d943e4059 100644
--- a/tests/python/unittest/test_gluon_contrib.py
+++ b/tests/python/unittest/test_gluon_contrib.py
@@ -25,7 +25,7 @@
     Concurrent, HybridConcurrent, Identity, SparseEmbedding, PixelShuffle1D,
     PixelShuffle2D, PixelShuffle3D)
 from mxnet.test_utils import almost_equal, default_context, assert_almost_equal, assert_allclose
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 import numpy as np
 
 
@@ -436,7 +436,3 @@ def test_ModulatedDeformableConvolution():
     with mx.autograd.record():
         y = net(x)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_data.py b/tests/python/unittest/test_gluon_data.py
index 29197bd58f6c..e6e3caebdf16 100644
--- a/tests/python/unittest/test_gluon_data.py
+++ b/tests/python/unittest/test_gluon_data.py
@@ -23,7 +23,7 @@
 import random
 from mxnet import gluon
 import platform
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.gluon.data import DataLoader
 import mxnet.ndarray as nd
 from mxnet import context
@@ -382,7 +382,3 @@ def test_dataloader_scope():
 
     assert item is not None
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_data_vision.py b/tests/python/unittest/test_gluon_data_vision.py
index b53dbf0687d3..d810f32b0afa 100644
--- a/tests/python/unittest/test_gluon_data_vision.py
+++ b/tests/python/unittest/test_gluon_data_vision.py
@@ -25,7 +25,7 @@
 from mxnet.gluon.data.vision import transforms
 from mxnet import image
 from mxnet.test_utils import *
-from common import assertRaises, setup_module, with_seed, teardown
+from common import assertRaises, setup_module, with_seed, teardown_module
 
 import numpy as np
 
@@ -42,12 +42,12 @@ def test_to_tensor():
     out_nd = transforms.ToTensor()(nd.array(data_in, dtype='uint8'))
     assert_almost_equal(out_nd.asnumpy(), np.transpose(
                         data_in.astype(dtype=np.float32) / 255.0, (0, 3, 1, 2)))
-    
+
     # Invalid Input
     invalid_data_in = nd.random.uniform(0, 255, (5, 5, 300, 300, 3)).astype(dtype=np.uint8)
     transformer = transforms.ToTensor()
     assertRaises(MXNetError, transformer, invalid_data_in)
-    
+
     # Bounds (0->0, 255->1)
     data_in = np.zeros((10, 20, 3)).astype(dtype=np.uint8)
     out_nd = transforms.ToTensor()(nd.array(data_in, dtype='uint8'))
@@ -126,7 +126,7 @@ def _test_resize_with_diff_type(dtype):
         assertRaises(MXNetError, invalid_transform, data_in)
 
     for dtype in ['uint8', 'float32', 'float64']:
-        _test_resize_with_diff_type(dtype)    
+        _test_resize_with_diff_type(dtype)
 
 
 @with_seed()
@@ -159,7 +159,7 @@ def _test_crop_resize_with_diff_type(dtype):
         # test with resize height and width should be greater than 0
         transformer = transforms.CropResize(0, 0, 100, 50, (-25, 25), 1)
         assertRaises(MXNetError, transformer, data_in)
-        # test height and width should be greater than 0 
+        # test height and width should be greater than 0
         transformer = transforms.CropResize(0, 0, -100, -50)
         assertRaises(MXNetError, transformer, data_in)
         # test cropped area is bigger than input data
@@ -168,7 +168,7 @@ def _test_crop_resize_with_diff_type(dtype):
         assertRaises(MXNetError, transformer, data_bath_in)
 
     for dtype in ['uint8', 'float32', 'float64']:
-        _test_crop_resize_with_diff_type(dtype)  
+        _test_crop_resize_with_diff_type(dtype)
 
     # test nd.image.crop backward
     def test_crop_backward(test_nd_arr, TestCase):
@@ -288,7 +288,7 @@ def test_random_rotation():
 @with_seed()
 def test_random_transforms():
     from mxnet.gluon.data.vision import transforms
-    
+
     tmp_t = transforms.Compose([transforms.Resize(300), transforms.RandomResizedCrop(224)])
     transform = transforms.Compose([transforms.RandomApply(tmp_t, 0.5)])
 
@@ -302,6 +302,3 @@ def test_random_transforms():
     assert_almost_equal(num_apply/float(iteration), 0.5, 0.1)
 
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_estimator.py b/tests/python/unittest/test_gluon_estimator.py
index 2c00b1609112..e33aa74b3ca7 100644
--- a/tests/python/unittest/test_gluon_estimator.py
+++ b/tests/python/unittest/test_gluon_estimator.py
@@ -20,13 +20,13 @@
 import sys
 import unittest
 import warnings
+import pytest
 
 import mxnet as mx
 from mxnet import gluon
 from mxnet.gluon import nn
 from mxnet.gluon.contrib.estimator import *
 from mxnet.gluon.contrib.estimator.event_handler import *
-from nose.tools import assert_raises
 
 
 def _get_test_network(params=None):
@@ -70,12 +70,12 @@ def test_fit():
     est.fit(train_data=dataloader,
             epochs=num_epochs)
 
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=dataiter,
                 epochs=num_epochs)
 
     # Input NDArray
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=[mx.nd.ones(shape=(10, 3))],
                 epochs=num_epochs)
 
@@ -107,12 +107,12 @@ def test_validation():
     val_metrics = est.val_metrics
     validation_handler = ValidationHandler(val_data=dataloader, eval_fn=est.evaluate)
 
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=dataiter,
                 val_data=dataiter,
                 epochs=num_epochs)
     # Input NDArray
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=[mx.nd.ones(shape=(10, 3))],
                 val_data=[mx.nd.ones(shape=(10, 3))],
                 epochs=num_epochs)
@@ -180,7 +180,7 @@ def test_trainer():
 
     # input invalid trainer
     trainer = 'sgd'
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est = Estimator(net=net,
                         loss=loss,
                         train_metrics=acc,
@@ -215,7 +215,7 @@ def test_metric():
     est.fit(train_data=train_data,
             epochs=num_epochs)
     # input invalid metric
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est = Estimator(net=net,
                         loss=loss,
                         train_metrics='acc',
@@ -238,7 +238,7 @@ def test_loss():
     net.initialize(ctx=ctx)
     trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.001})
     # input invalid loss
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est = Estimator(net=net,
                         loss='mse',
                         train_metrics=acc,
@@ -264,13 +264,13 @@ def test_context():
                     train_metrics=metrics,
                     context=ctx)
     # input invalid context
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est = Estimator(net=net,
                         loss=loss,
                         train_metrics=metrics,
                         context='cpu')
 
-    with assert_raises(AssertionError):
+    with pytest.raises(AssertionError):
         est = Estimator(net=net,
                         loss=loss,
                         train_metrics=metrics,
@@ -360,7 +360,7 @@ def test_default_handlers():
     # handler with mixed metrics, some handler use metrics prepared by estimator
     # some handler use metrics user prepared
     logging = LoggingHandler(metrics=[mx.metric.RMSE("val acc")])
-    with assert_raises(ValueError):
+    with pytest.raises(ValueError):
         est.fit(train_data=train_data, epochs=num_epochs, event_handlers=[logging])
 
     # test handler order
@@ -394,7 +394,7 @@ def test_val_net():
                     val_loss=val_loss,
                     val_net=val_net)
 
-    with assert_raises(RuntimeError):
+    with pytest.raises(RuntimeError):
         est.fit(train_data=dataloader,
                 val_data=dataloader,
                 epochs=num_epochs)
diff --git a/tests/python/unittest/test_gluon_model_zoo.py b/tests/python/unittest/test_gluon_model_zoo.py
index d53dd403a5b8..a921d3de18e6 100644
--- a/tests/python/unittest/test_gluon_model_zoo.py
+++ b/tests/python/unittest/test_gluon_model_zoo.py
@@ -19,7 +19,7 @@
 import mxnet as mx
 from mxnet.gluon.model_zoo.vision import get_model
 import sys
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 import multiprocessing
 
 
@@ -66,6 +66,3 @@ def test_parallel_download():
     for p in processes:
         p.join()
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_rnn.py b/tests/python/unittest/test_gluon_rnn.py
index f3ffd03f5dbb..4e8db2fbcab4 100644
--- a/tests/python/unittest/test_gluon_rnn.py
+++ b/tests/python/unittest/test_gluon_rnn.py
@@ -842,7 +842,3 @@ def hybrid_forward(self, F, inputs, valid_len):
     _check_bidirectional_unroll_valid_length(1)
     _check_bidirectional_unroll_valid_length(3)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_gluon_trainer.py b/tests/python/unittest/test_gluon_trainer.py
index 350700cc129f..414b15c3b3f2 100644
--- a/tests/python/unittest/test_gluon_trainer.py
+++ b/tests/python/unittest/test_gluon_trainer.py
@@ -24,7 +24,7 @@
 from mxnet.test_utils import assert_almost_equal
 from common import setup_module, with_seed, assertRaises
 from copy import deepcopy
-from nose.tools import raises, assert_raises
+import pytest
 
 def dict_equ(a, b):
     assert set(a) == set(b)
@@ -32,7 +32,7 @@ def dict_equ(a, b):
         assert (a[k].asnumpy() == b[k].asnumpy()).all()
 
 @with_seed()
-@raises(RuntimeError)
+@pytest.mark.xfail(raises=RuntimeError)
 def test_multi_trainer():
     x = gluon.Parameter('x', shape=(10,), stype='row_sparse')
     x.initialize()
@@ -78,7 +78,7 @@ def test_trainer_with_teststore():
     # Expect exceptions if update_on_kvstore is set to True,
     # because TestStore does not support that
     invalid_trainer = gluon.Trainer([x], 'sgd', kvstore=kv, update_on_kvstore=True)
-    assert_raises(ValueError, invalid_trainer._init_kvstore)
+    pytest.raises(ValueError, invalid_trainer._init_kvstore)
 
 @with_seed()
 def test_trainer():
@@ -110,8 +110,8 @@ def test_trainer():
         dict_equ(trainer._kvstore._updater.states, states)
         assert trainer._optimizer == trainer._kvstore._updater.optimizer
         # invalid usage of update and allreduce_grads if update_on_kvstore
-        assert_raises(AssertionError, trainer.update, 1)
-        assert_raises(AssertionError, trainer.allreduce_grads)
+        pytest.raises(AssertionError, trainer.update, 1)
+        pytest.raises(AssertionError, trainer.allreduce_grads)
     else:
         for updater in trainer._updaters:
             dict_equ(updater.states, states)
diff --git a/tests/python/unittest/test_gluon_utils.py b/tests/python/unittest/test_gluon_utils.py
index bc816b1794ee..4e37596fab0b 100644
--- a/tests/python/unittest/test_gluon_utils.py
+++ b/tests/python/unittest/test_gluon_utils.py
@@ -29,7 +29,7 @@
     import mock
 import mxnet as mx
 import requests
-from nose.tools import raises
+import pytest
 
 
 class MockResponse(requests.Response):
@@ -40,7 +40,7 @@ def __init__(self, status_code, content):
         self.raw = io.BytesIO(content.encode('utf-8'))
 
 
-@raises(Exception)
+@pytest.mark.xfail(raises=Exception)
 @mock.patch(
     'requests.get', mock.Mock(side_effect=requests.exceptions.ConnectionError))
 def test_download_retries():
diff --git a/tests/python/unittest/test_higher_order_grad.py b/tests/python/unittest/test_higher_order_grad.py
index ad31c34bd590..28357360bc41 100644
--- a/tests/python/unittest/test_higher_order_grad.py
+++ b/tests/python/unittest/test_higher_order_grad.py
@@ -22,8 +22,6 @@
 from operator import mul
 import random
 
-from nose.tools import ok_
-
 from common import with_seed
 import mxnet
 from mxnet import nd, autograd, gluon
@@ -648,18 +646,18 @@ def test_dense_backward_flatten():
         w_grad_grad_e = nd.dot(o_y, o_x_grad, transpose_a=True)
         x_grad_e = nd.dot(o_y, w)
         x_grad_grad_e = nd.dot(o_y, o_w_grad)
-        ok_(w_grad.shape == w.shape)
-        ok_(w_grad_grad.shape == w.shape)
-        ok_(x_grad.shape == x.shape)
-        ok_(x_grad_grad.shape == x.shape)
+        assert w_grad.shape == w.shape
+        assert w_grad_grad.shape == w.shape
+        assert x_grad.shape == x.shape
+        assert x_grad_grad.shape == x.shape
         w_grad_check = same(flatten2d_right(w_grad), flatten2d_right(w_grad_e))
         w_grad_grad_check = same(flatten2d_right(w_grad_grad), flatten2d_right(w_grad_grad_e))
         x_grad_check = same(flatten2d_right(x_grad), flatten2d_right(x_grad_e))
         x_grad_grad_check = same(flatten2d_right(x_grad_grad), flatten2d_right(x_grad_grad_e))
-        ok_(x_grad_check)
-        ok_(w_grad_check)
-        ok_(x_grad_grad_check)
-        ok_(w_grad_grad_check)
+        assert x_grad_check
+        assert w_grad_check
+        assert x_grad_grad_check
+        assert w_grad_grad_check
 
 @with_seed()
 def test_dense_backward_no_flatten():
@@ -701,12 +699,8 @@ def test_dense_backward_no_flatten():
         w_grad_grad_check = same(flatten2d_left(w_grad_grad), flatten2d_left(w_grad_grad_e))
         x_grad_check = same(flatten2d_left(x_grad), flatten2d_left(x_grad_e))
         x_grad_grad_check = same(flatten2d_left(x_grad_grad), flatten2d_left(x_grad_grad_e))
-        ok_(x_grad_check)
-        ok_(w_grad_check)
-        ok_(x_grad_grad_check)
-        ok_(w_grad_grad_check)
-
+        assert x_grad_check
+        assert w_grad_check
+        assert x_grad_grad_check
+        assert w_grad_grad_check
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_image.py b/tests/python/unittest/test_image.py
index 033b8e5aab04..4a527948cb6f 100644
--- a/tests/python/unittest/test_image.py
+++ b/tests/python/unittest/test_image.py
@@ -23,8 +23,7 @@
 import shutil
 import tempfile
 import unittest
-
-from nose.tools import raises
+import pytest
 
 
 def _get_data(url, dirname):
@@ -126,7 +125,7 @@ def teardownClass(cls):
             print("cleanup {}".format(cls.IMAGES_DIR))
             shutil.rmtree(cls.IMAGES_DIR)
 
-    @raises(mx.base.MXNetError)
+    @pytest.mark.xfail(raises=mx.base.MXNetError)
     def test_imread_not_found(self):
         x = mx.img.image.imread("/139810923jadjsajlskd.___adskj/blah.jpg")
 
@@ -162,11 +161,11 @@ def test_imdecode_bytearray(self):
             cv_image = cv2.imread(img)
             assert_almost_equal(image.asnumpy(), cv_image)
 
-    @raises(mx.base.MXNetError)
+    @pytest.mark.xfail(raises=mx.base.MXNetError)
     def test_imdecode_empty_buffer(self):
         mx.image.imdecode(b'', to_rgb=0)
 
-    @raises(mx.base.MXNetError)
+    @pytest.mark.xfail(raises=mx.base.MXNetError)
     def test_imdecode_invalid_image(self):
         image = mx.image.imdecode(b'clearly not image content')
         assert_equal(image, None)
@@ -451,7 +450,3 @@ def test_random_rotate(self):
                                                  angle_limits)
         self.assertEqual(out_batch_image.shape, (3, 3, 30, 60))
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_infer_shape.py b/tests/python/unittest/test_infer_shape.py
index 1312be0c0081..a2f4cfef3701 100644
--- a/tests/python/unittest/test_infer_shape.py
+++ b/tests/python/unittest/test_infer_shape.py
@@ -18,7 +18,7 @@
 # pylint: skip-file
 import mxnet as mx
 from common import models
-from nose.tools import *
+import pytest
 
 def test_mlp2_infer_shape():
     # Build MLP
@@ -36,7 +36,7 @@ def test_mlp2_infer_shape():
     for k, v in true_shapes.items():
         assert arg_shape_dict[k] == v
 
-@raises(mx.MXNetError)
+@pytest.mark.xfail(raises=mx.MXNetError)
 def test_mlp2_infer_error():
     # Test shape inconsistent case
     out = models.mlp2()
@@ -246,6 +246,3 @@ def test_where_partial_shape():
         _, result, _ = where_op.infer_shape_partial(cond=(-1,), x=(2, 2), y=(2, 2))
         assert result == [None]
 
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_infer_type.py b/tests/python/unittest/test_infer_type.py
index bad83f3ef01b..286556a006c4 100644
--- a/tests/python/unittest/test_infer_type.py
+++ b/tests/python/unittest/test_infer_type.py
@@ -20,7 +20,6 @@
 import numpy as np
 from common import models, with_seed
 from mxnet import autograd
-from nose.tools import *
 from mxnet.test_utils import assert_almost_equal
 
 @with_seed()
@@ -52,7 +51,3 @@ def test_func(a):
         test64.backward()
     assert_almost_equal(data64.grad.asnumpy(), data32.grad.asnumpy(), atol=1e-5, rtol=1e-5)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
\ No newline at end of file
diff --git a/tests/python/unittest/test_io.py b/tests/python/unittest/test_io.py
index a13addb0adca..a6e2cb19d21f 100644
--- a/tests/python/unittest/test_io.py
+++ b/tests/python/unittest/test_io.py
@@ -149,7 +149,7 @@ def _test_last_batch_handle(data, labels=None, is_image=False):
         batch_count_list = [40, 39, 39]
     else:
         batch_count_list = [8, 7, 7]
-    
+
     for idx in range(len(last_batch_handle_list)):
         dataiter = mx.io.NDArrayIter(
             data, labels, 128, False, last_batch_handle=last_batch_handle_list[idx])
@@ -168,7 +168,7 @@ def _test_last_batch_handle(data, labels=None, is_image=False):
                        labelcount[int(label[i])] += 1
             else:
                 assert not batch.label, 'label is not empty list'
-            # keep the last batch of 'pad' to be used later 
+            # keep the last batch of 'pad' to be used later
             # to test first batch of roll_over in second iteration
             batch_count += 1
             if last_batch_handle_list[idx] == 'pad' and \
@@ -243,7 +243,7 @@ def test_NDArrayIter_h5py():
     with h5py.File('ndarraytest.h5') as f:
         f.create_dataset('data', data=data)
         f.create_dataset('label', data=labels)
-        
+
         _test_last_batch_handle(f['data'], f['label'])
         _test_last_batch_handle(f['data'], [])
         _test_last_batch_handle(f['data'])
@@ -285,7 +285,7 @@ def test_NDArrayIter_csr():
                      {'data': train_data}, dns, batch_size)
     except ImportError:
         pass
-    
+
     # scipy.sparse.csr_matrix with shuffle
     csr_iter = iter(mx.io.NDArrayIter({'data': train_data}, dns, batch_size,
                                       shuffle=True, last_batch_handle='discard'))
@@ -411,16 +411,15 @@ def check_libSVMIter_exception():
 
 
 def test_DataBatch():
-    from nose.tools import ok_
     from mxnet.io import DataBatch
     import re
     batch = DataBatch(data=[mx.nd.ones((2, 3))])
-    ok_(re.match(
-        'DataBatch: data shapes: \[\(2L?, 3L?\)\] label shapes: None', str(batch)))
+    assert re.match(
+        'DataBatch: data shapes: \[\(2L?, 3L?\)\] label shapes: None', str(batch))
     batch = DataBatch(data=[mx.nd.ones((2, 3)), mx.nd.ones(
         (7, 8))], label=[mx.nd.ones((4, 5))])
-    ok_(re.match(
-        'DataBatch: data shapes: \[\(2L?, 3L?\), \(7L?, 8L?\)\] label shapes: \[\(4L?, 5L?\)\]', str(batch)))
+    assert re.match(
+        'DataBatch: data shapes: \[\(2L?, 3L?\), \(7L?, 8L?\)\] label shapes: \[\(4L?, 5L?\)\]', str(batch))
 
 
 def test_CSVIter():
@@ -462,7 +461,7 @@ def assert_dataiter_items_equals(dataiter1, dataiter2):
         are the equal.
         """
         for batch1, batch2 in zip_longest(dataiter1, dataiter2):
-            
+
             # ensure iterators contain the same number of batches
             # zip_longest will return None if on of the iterators have run out of batches
             assert batch1 and batch2, 'The iterators do not contain the same number of batches'
@@ -533,7 +532,7 @@ def assert_dataiter_items_not_equals(dataiter1, dataiter2):
         random_h=10,
         max_shear_ratio=2,
         seed_aug=seed_aug)
-    
+
     assert_dataiter_items_equals(dataiter1, dataiter2)
 
     # check whether to get different images after change seed_aug
@@ -573,7 +572,7 @@ def assert_dataiter_items_not_equals(dataiter1, dataiter2):
         data_shape=(3, 28, 28),
         batch_size=3,
         seed_aug=seed_aug)
-    
+
     assert_dataiter_items_equals(dataiter1, dataiter2)
 
 if __name__ == "__main__":
diff --git a/tests/python/unittest/test_kvstore.py b/tests/python/unittest/test_kvstore.py
index 28d4ec262c06..a3d54469b28e 100644
--- a/tests/python/unittest/test_kvstore.py
+++ b/tests/python/unittest/test_kvstore.py
@@ -20,7 +20,7 @@
 import numpy as np
 import unittest
 from mxnet.test_utils import rand_ndarray, assert_almost_equal
-from common import setup_module, with_seed, assertRaises, teardown
+from common import setup_module, with_seed, assertRaises, teardown_module
 from mxnet.base import py_str, MXNetError
 
 shape = (4, 4)
@@ -344,6 +344,3 @@ def check_invalid_key_types_list(kv, key):
         check_invalid_key_types_single(kvs[i], single_keys[1 - i])
         check_invalid_key_types_list(kvs[i], list_keys[1 - i])
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_kvstore_custom.py b/tests/python/unittest/test_kvstore_custom.py
index 4f1f309d24c1..a8f0869f0236 100644
--- a/tests/python/unittest/test_kvstore_custom.py
+++ b/tests/python/unittest/test_kvstore_custom.py
@@ -20,7 +20,7 @@
 import numpy as np
 import unittest
 from mxnet.test_utils import rand_ndarray, assert_almost_equal
-from common import setup_module, with_seed, assertRaises, teardown
+from common import setup_module, with_seed, assertRaises, teardown_module
 from mxnet.base import py_str, MXNetError
 
 shape = (4, 4)
@@ -190,6 +190,3 @@ def check_unsupported_methods(kv):
     kv = mx.kv.create('teststore')
     check_unsupported_methods(kv)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_loss.py b/tests/python/unittest/test_loss.py
index a1a49c97d7f4..e779fd672701 100644
--- a/tests/python/unittest/test_loss.py
+++ b/tests/python/unittest/test_loss.py
@@ -19,7 +19,7 @@
 import numpy as np
 from mxnet import gluon, autograd
 from mxnet.test_utils import assert_almost_equal, default_context
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 import unittest
 
 
@@ -354,7 +354,7 @@ def test_sdml_loss():
     N = 5 # number of samples
     DIM = 10 # Dimensionality
     EPOCHS = 20
-    
+
     # Generate randomized data and 'positive' samples
     data = mx.random.uniform(-1, 1, shape=(N, DIM))
     pos = data + mx.random.uniform(-0.1, 0.1, shape=(N, DIM)) # correlated paired data
@@ -380,7 +380,7 @@ def test_sdml_loss():
     # After training euclidean distance between aligned pairs should be lower than all non-aligned pairs
     avg_loss = loss.sum()/len(loss)
     assert(avg_loss < 0.05)
-    
+
 @with_seed()
 def test_cosine_loss():
     #Generating samples
@@ -488,7 +488,3 @@ def test_bce_loss_with_pos_weight():
     npy_bce_loss = (- label_npy * np.log(prob_npy)*pos_weight_npy - (1 - label_npy) * np.log(1 - prob_npy)).mean(axis=1)
     assert_almost_equal(mx_bce_loss, npy_bce_loss, rtol=1e-4, atol=1e-5)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_metric.py b/tests/python/unittest/test_metric.py
index e7273fba35d5..d1e1c5a35fb3 100644
--- a/tests/python/unittest/test_metric.py
+++ b/tests/python/unittest/test_metric.py
@@ -407,7 +407,3 @@ def test_single_array_input():
     rmse.get()
     _, rmse_res = rmse.get()
     np.testing.assert_almost_equal(rmse_res, 0.1)
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_metric_perf.py b/tests/python/unittest/test_metric_perf.py
index 36cbc685797c..fc0f8da5d451 100644
--- a/tests/python/unittest/test_metric_perf.py
+++ b/tests/python/unittest/test_metric_perf.py
@@ -118,7 +118,3 @@ def test_metric_performance():
                     run_metric(k, v[1], (data_size * 128)//(n * c), n, c, pred_ctx, label_ctx, **v[0])
                 print("{:-^90}".format(''), file=sys.stderr)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_module.py b/tests/python/unittest/test_module.py
index b82933126d67..65d86f62baf4 100644
--- a/tests/python/unittest/test_module.py
+++ b/tests/python/unittest/test_module.py
@@ -22,7 +22,7 @@
 import numpy as np
 from functools import reduce
 from mxnet.module.executor_group import DataParallelExecutorGroup
-from common import setup_module, with_seed, assertRaises, teardown
+from common import setup_module, with_seed, assertRaises, teardown_module
 from collections import namedtuple
 curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
 sys.path.insert(0, os.path.join(curr_path, "../train"))
@@ -1029,7 +1029,3 @@ def get_module_idx2name(mod):
     mod2.init_optimizer(optimizer=opt)
     assert mod2._optimizer.idx2name == get_module_idx2name(mod2)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_ndarray.py b/tests/python/unittest/test_ndarray.py
index 3a9bd9e93126..22caf77536d1 100644
--- a/tests/python/unittest/test_ndarray.py
+++ b/tests/python/unittest/test_ndarray.py
@@ -23,7 +23,7 @@
 import pickle as pkl
 import random
 import functools
-from nose.tools import assert_raises, raises
+import pytest
 from common import with_seed, assertRaises, TemporaryDirectory
 from mxnet.test_utils import almost_equal
 from mxnet.test_utils import assert_almost_equal, assert_exception
@@ -1341,7 +1341,7 @@ def check_fluent_regular(func, kwargs, shape=(5, 17, 1), equal_nan=False):
     check_fluent_regular('squeeze', {'axis': (1, 3)}, shape=(2, 1, 3, 1, 4))
 
 
-@raises(ValueError)
+@pytest.mark.xfail(raises=ValueError)
 def test_bool_ambiguous():
     bool(mx.nd.ones((2,3,4)))
 
@@ -1603,10 +1603,10 @@ def convert(num):
                   # Test basic indexing with newaxis
                   (None, False),
                   ((1, None, -2, 3, -4), False),
-                  ((1, slice(2, 5), None), False), 
-                  ((slice(None), slice(1, 4), None, slice(2, 3)), False), 
-                  ((slice(1, 3), slice(1, 3), slice(1, 3), slice(1, 3), None), False), 
-                  ((slice(1, 3), slice(1, 3), None, slice(1, 3), slice(1, 3)), False), 
+                  ((1, slice(2, 5), None), False),
+                  ((slice(None), slice(1, 4), None, slice(2, 3)), False),
+                  ((slice(1, 3), slice(1, 3), slice(1, 3), slice(1, 3), None), False),
+                  ((slice(1, 3), slice(1, 3), None, slice(1, 3), slice(1, 3)), False),
                   ((None, slice(1, 2), 3, None), False),
                   ((1, None, 2, 3, None, None, 4), False),
                   # Advanced indexing
@@ -2062,7 +2062,3 @@ def test_load_saved_gpu_array_when_no_gpus_are_present():
     # but there are no GPUs
     array.__setstate__(ndarray_state)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_numpy_gluon.py b/tests/python/unittest/test_numpy_gluon.py
index f3f01fc2ae92..713591795a43 100644
--- a/tests/python/unittest/test_numpy_gluon.py
+++ b/tests/python/unittest/test_numpy_gluon.py
@@ -431,7 +431,3 @@ def hybrid_forward(self, F, valid_length):
 
     assert mx.test_utils.same(out1.asnumpy(), out2.asnumpy())
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_numpy_interoperability.py b/tests/python/unittest/test_numpy_interoperability.py
index 1f795721b820..9a737f4d80b8 100644
--- a/tests/python/unittest/test_numpy_interoperability.py
+++ b/tests/python/unittest/test_numpy_interoperability.py
@@ -1760,7 +1760,7 @@ def test_shapes():
             a = np.ones((2,), dtype=dt)
             b = np.ones((2,), dtype=dt)
             OpArgMngr.add_workload('matmul', a, b)
-    
+
     def test_result_types():
         mat = np.ones((1,1))
         vec = np.ones((1,))
@@ -1769,7 +1769,7 @@ def test_result_types():
             v = vec.astype(dt)
             for arg in [(m, v), (v, m), (m, m)]:
                 OpArgMngr.add_workload('matmul', *arg)
-    
+
     def test_scalar_output():
         vec1 = np.array([2])
         vec2 = np.array([3, 4]).reshape(1, -1)
@@ -1778,7 +1778,7 @@ def test_scalar_output():
             v2 = vec2.astype(dt)
             OpArgMngr.add_workload('matmul', v1, v2)
             OpArgMngr.add_workload('matmul', v2.T, v1)
-    
+
     def test_vector_vector_values():
         vec1 = np.array([1, 2])
         vec2 = np.array([3, 4]).reshape(-1, 1)
@@ -1810,7 +1810,7 @@ def test_matrix_vector_values():
             m2 = mat2.astype(dt)
             OpArgMngr.add_workload('matmul', m1, v)
             OpArgMngr.add_workload('matmul', m2, v)
-    
+
     def test_matrix_matrix_values():
         mat1 = np.array([[1, 2], [3, 4]])
         mat2 = np.array([[1, 0], [1, 1]])
@@ -3174,7 +3174,3 @@ def test_np_fallback_ops():
     op_list = np.fallback.__all__ + ['linalg.{}'.format(op_name) for op_name in np.fallback_linalg.__all__]
     check_interoperability(op_list)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_numpy_ndarray.py b/tests/python/unittest/test_numpy_ndarray.py
index 3ce53c6a6e80..a45ce408802b 100644
--- a/tests/python/unittest/test_numpy_ndarray.py
+++ b/tests/python/unittest/test_numpy_ndarray.py
@@ -623,7 +623,7 @@ def test_nd_no_format():
     if str(context)[:3] != 'gpu':
         test_0d()
         test_nd_format()
-        test_nd_no_format() 
+        test_nd_no_format()
     # if the program is running in GPU, the formatted string would be appended with context notation
     # for exmpale, if a = np.array([np.pi]), the return value of '{}'.format(a) is '[3.1415927] @gpu(0)'
 
@@ -1237,7 +1237,7 @@ def test_boolean_indexing_assign():
 
         mx_mask = np.array([[False,True, True],[False, True,False]],dtype=np.bool)
         np_mask = mx_mask.asnumpy()
-        
+
         np_data[0, np_mask] = 5
         mx_data[0, mx_mask] = 5
         assert_almost_equal(mx_data.asnumpy(), np_data, rtol=1e-3, atol=1e-5, use_broadcast=False)
@@ -1315,7 +1315,3 @@ def test_np_ndarray_pickle():
             a_load = pickle.load(f)
         same(a.asnumpy(), a_load.asnumpy())
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_numpy_op.py b/tests/python/unittest/test_numpy_op.py
index 063c3b7a58a6..ac2eb48f8446 100644
--- a/tests/python/unittest/test_numpy_op.py
+++ b/tests/python/unittest/test_numpy_op.py
@@ -26,7 +26,7 @@
 import mxnet as mx
 import scipy.stats as ss
 import scipy.special as scipy_special
-from nose.tools import assert_raises
+import pytest
 from mxnet import np, npx
 from mxnet.gluon import HybridBlock
 from mxnet.base import MXNetError
@@ -552,7 +552,7 @@ def ShapeReduce(mat, shape, is_b=False):
     for shape_a, shape_b in bad_shapes:
         a = np.random.uniform(size=shape_a)
         b = np.random.uniform(size=shape_b)
-        assert_raises(MXNetError, lambda: np.matmul(a, b))
+        pytest.raises(MXNetError, lambda: np.matmul(a, b))
 
 
 @with_seed()
@@ -1478,7 +1478,7 @@ def gt_grad_batch_dot_numpy(lhs, rhs, ograd, transpose_a, transpose_b, lhs_req,
         for dtype in dtypes:
             lhs_val = mx.np.array(_np.random.uniform(-1.0, 1.0, lhs_shape), dtype=dtype)
             rhs_val = mx.np.array(_np.random.uniform(-1.0, 1.0, rhs_shape), dtype=dtype)
-            assert_raises(MXNetError, lambda: mx.npx.batch_dot(lhs_val, rhs_val,
+            pytest.raises(MXNetError, lambda: mx.npx.batch_dot(lhs_val, rhs_val,
                                                                transpose_a=transpose_a,
                                                                transpose_b=transpose_b))
 
@@ -1962,8 +1962,8 @@ def hybrid_forward(self, F, a):
                             assert_almost_equal(mx_out.asnumpy(), np_out, rtol=1e-3, atol=1e-5, use_broadcast=False)
     # Test for error raising
     dat = np.random.normal(0, 1, (3, 4, 5), dtype=np.float32)
-    assert_raises(ValueError, lambda: dat.transpose((0, 0, 1)))
-    assert_raises(MXNetError, lambda: dat.transpose((0, 1, 3)))
+    pytest.raises(ValueError, lambda: dat.transpose((0, 0, 1)))
+    pytest.raises(MXNetError, lambda: dat.transpose((0, 1, 3)))
 
 
 
@@ -5556,7 +5556,7 @@ class TestLstsq(HybridBlock):
         def __init__(self, rcond):
             super(TestLstsq, self).__init__()
             self._rcond = rcond
-        
+
         def hybrid_forward(self, F, a, b, rcond='warn'):
             return F.np.linalg.lstsq(a, b, rcond=self._rcond)
 
@@ -6876,10 +6876,10 @@ def __init__(self, n, k=0, m=None):
             if m is None:
                 m = n
             self._m = m
-        
+
         def hybrid_forward(self, F, x, *args, **kwargs):
             return x, F.np.tril_indices(n=self._n, k=self._k, m=self._m)
-    
+
     for n in _np.random.random_integers(-10, 50, 2):
         for k in _np.random.random_integers(-50, 50, 2):
             for m in _np.random.random_integers(-10, 50, 2):
@@ -6900,7 +6900,7 @@ def hybrid_forward(self, F, x, *args, **kwargs):
                         np_data[np_out] = -10
                         mx_data[mx_out] = -10
                         assert same(np_data, mx_data.asnumpy())
-                        
+
 
 @with_seed()
 @use_np
@@ -7614,7 +7614,7 @@ def hybrid_forward(self, F, a):
         a = np.random.uniform(-1.0, 1.0, size=a_shape)
         np_out = _np.median(a.asnumpy(), axis=axis, keepdims=keepdims)
         mx_out = test_median(a)
-        
+
         assert mx_out.shape == np_out.shape
         assert_almost_equal(mx_out.asnumpy(), np_out, atol=atol, rtol=rtol)
 
@@ -8630,10 +8630,10 @@ def __init__(self, left=None, right=None, period=None):
             self._left = left
             self._right = right
             self._period = period
-        
+
         def hybrid_forward(self, F, x, xp, fp):
             return F.np.interp(x, xp, fp, left=self._left, right=self._right, period=self._period)
-    
+
     class TestInterpScalar(HybridBlock):
         def __init__(self, x=None, left=None, right=None, period=None):
             super(TestInterpScalar, self).__init__()
@@ -8641,7 +8641,7 @@ def __init__(self, x=None, left=None, right=None, period=None):
             self._left = left
             self._right = right
             self._period = period
-        
+
         def hybrid_forward(self, F, xp, fp):
             return F.np.interp(self._x, xp, fp, left=self._left, right=self._right, period=self._period)
 
@@ -8668,13 +8668,13 @@ def hybrid_forward(self, F, xp, fp):
         else:
             x = np.random.uniform(0, 100, size=xshape).astype(xtype)
             xp = np.sort(np.random.choice(100, dsize, replace=False).astype(dtype))
-            fp = np.random.uniform(-50, 50, size=dsize).astype(dtype) 
+            fp = np.random.uniform(-50, 50, size=dsize).astype(dtype)
         np_x = x.asnumpy()
         if x_scalar and xshape == ():
             x = x.item()
             np_x = x
             test_interp = TestInterpScalar(x=x, left=left, right=right, period=period)
-        else: 
+        else:
             test_interp = TestInterp(left=left, right=right, period=period)
         if hybridize:
             test_interp.hybridize()
@@ -8804,7 +8804,7 @@ def __init__(self, axis=0, start=0):
             super(TestRollaxis, self).__init__()
             self._axis = axis
             self._start = start
-             
+
         def hybrid_forward(self, F, a, *args, **kwargs):
             return F.np.rollaxis(a, axis=self._axis, start=self._start)
 
@@ -8835,7 +8835,3 @@ def hybrid_forward(self, F, a, *args, **kwargs):
                         assert np_out.dtype == mx_out.dtype
                         assert same(mx_out.asnumpy(), np_out)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_operator.py b/tests/python/unittest/test_operator.py
index 0c795db9e381..a1d467eff319 100644
--- a/tests/python/unittest/test_operator.py
+++ b/tests/python/unittest/test_operator.py
@@ -29,9 +29,9 @@
 from mxnet.test_utils import *
 from mxnet.operator import *
 from mxnet.base import py_str, MXNetError, _as_list
-from common import setup_module, with_seed, teardown, assert_raises_cudnn_not_satisfied, assert_raises_cuda_not_satisfied, assertRaises
+from common import setup_module, with_seed, teardown_module, assert_raises_cudnn_not_satisfied, assert_raises_cuda_not_satisfied, assertRaises
 from common import run_in_spawned_process
-from nose.tools import assert_raises, ok_
+import pytest
 import unittest
 import os
 
@@ -6069,7 +6069,7 @@ def f(in_data, out_data):
         b = mx.nd.zeros((1, 4))
         c = mx.nd.Custom(a, b, op_type='Dot1')
         c.wait_to_read()
-    assert_raises(MXNetError, custom_exc1)
+    pytest.raises(MXNetError, custom_exc1)
 
     # 2. error in pushing operator to engine
     def custom_exc2():
@@ -6081,7 +6081,7 @@ def f(in_data, out_data):
         # trigger error by invalid input shapes of operands
         c = mx.nd.Custom(a, b, op_type='Dot2')
         c.wait_to_read()
-    assert_raises(MXNetError, custom_exc2)
+    pytest.raises(MXNetError, custom_exc2)
 
     # 3. error in real execution
     if default_context().device_type == 'cpu':
@@ -6098,7 +6098,7 @@ def f(in_data, out_data):
             b = mx.nd.zeros((1, 2))
             c = mx.nd.Custom(a, b, op_type='Dot3')
             c.wait_to_read()
-        assert_raises(MXNetError, custom_exc3)
+        pytest.raises(MXNetError, custom_exc3)
 
         def custom_exc4():
             def f(in_data, out_data):
@@ -6112,7 +6112,7 @@ def f(in_data, out_data):
             b = mx.nd.zeros((1, 2))
             c = mx.nd.Custom(a, b, op_type='Dot4')
             c.wait_to_read()
-        assert_raises(MXNetError, custom_exc4)
+        pytest.raises(MXNetError, custom_exc4)
 
 
 @with_seed()
@@ -7157,11 +7157,11 @@ def test_dropout_reproducibility():
 
     assert_almost_equal(result1.asnumpy(), result5.asnumpy())
     assert_almost_equal(result2.asnumpy(), result6.asnumpy())
-    with assert_raises(AssertionError):
+    with pytest.raises(AssertionError):
         assert_almost_equal(result1.asnumpy(), result2.asnumpy())
-    with assert_raises(AssertionError):
+    with pytest.raises(AssertionError):
         assert_almost_equal(result1.asnumpy(), result3.asnumpy())
-    with assert_raises(AssertionError):
+    with pytest.raises(AssertionError):
         assert_almost_equal(result2.asnumpy(), result4.asnumpy())
 
 @unittest.skip("test fails intermittently. temporarily disabled till it gets fixed. tracked at https://github.com/apache/incubator-mxnet/issues/11290")
@@ -7704,8 +7704,8 @@ def max():
         a = mx.nd.zeros(shape=(5, 0))
         a.max()
 
-    assert_raises(MXNetError, min)
-    assert_raises(MXNetError, max)
+    pytest.raises(MXNetError, min)
+    pytest.raises(MXNetError, max)
 
 
 @with_seed()
@@ -9330,18 +9330,18 @@ def test_add_n():
 
 def test_get_all_registered_operators():
     ops = get_all_registered_operators()
-    ok_(isinstance(ops, list))
-    ok_(len(ops) > 0)
-    ok_('Activation' in ops)
+    assert isinstance(ops, list)
+    assert len(ops) > 0
+    assert 'Activation' in ops
 
 
 def test_get_operator_arguments():
     operator_arguments = get_operator_arguments('Activation')
-    ok_(isinstance(operator_arguments, OperatorArguments))
-    ok_(operator_arguments.names == ['data', 'act_type'])
-    ok_(operator_arguments.types
-        == ['NDArray-or-Symbol', "{'relu', 'sigmoid', 'softrelu', 'softsign', 'tanh'}, required"])
-    ok_(operator_arguments.narg == 2)
+    assert isinstance(operator_arguments, OperatorArguments)
+    assert operator_arguments.names == ['data', 'act_type']
+    assert operator_arguments.types \
+        == ['NDArray-or-Symbol', "{'relu', 'sigmoid', 'softrelu', 'softsign', 'tanh'}, required"]
+    assert operator_arguments.narg == 2
 
 
 def test_transpose_infer_shape_back():
@@ -9924,7 +9924,3 @@ def test_elemwise_sum_for_gradient_accumulation():
         assert stored_grad['write'] == stored_grad['add']
         assert stored_grad['write'] == 2 * nrepeat
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_optimizer.py b/tests/python/unittest/test_optimizer.py
old mode 100755
new mode 100644
index 2a15e3407862..4488d5a62e45
--- a/tests/python/unittest/test_optimizer.py
+++ b/tests/python/unittest/test_optimizer.py
@@ -22,10 +22,10 @@
 import mxnet.lr_scheduler as lr_scheduler
 from mxnet import gluon
 import unittest
-from nose.tools import raises
+import pytest
 import math
 from mxnet.test_utils import *
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 @with_seed()
 def test_learning_rate():
@@ -44,7 +44,7 @@ def test_learning_rate():
     assert o3.learning_rate == 1024
 
 
-@raises(UserWarning)
+@pytest.mark.xfail(raises=UserWarning)
 @with_seed()
 def test_learning_rate_expect_user_warning():
     lr_s = lr_scheduler.FactorScheduler(step=1)
@@ -284,7 +284,7 @@ def test_lars():
 def test_lamb():
     opt1 = mx.optimizer.LAMB
     opt2 = mx.optimizer.LAMB
-    
+
     shapes = [(3, 4, 5), (10, 4), (7,)]
     beta1_options = [{}, {'beta1': 0.5}]
     beta2_options = [{}, {'beta2': 0.8}]
@@ -952,8 +952,3 @@ def test_cosine_scheduler():
     np.testing.assert_almost_equal(cosine_sched(steps), final_lr)
     assert (cosine_sched(500) > 1.5)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
-
diff --git a/tests/python/unittest/test_predictor.py b/tests/python/unittest/test_predictor.py
index fc2fbf600cbc..325b830e4226 100644
--- a/tests/python/unittest/test_predictor.py
+++ b/tests/python/unittest/test_predictor.py
@@ -26,7 +26,7 @@
 import mxnet.ndarray as nd
 from mxnet import gluon
 from mxnet.test_utils import assert_almost_equal
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 @with_seed()
 def test_predictor():
@@ -81,7 +81,3 @@ def test_load_ndarray():
     for k in nd_data.keys():
         assert_almost_equal(nd_data[k].asnumpy(), nd_load[k], rtol=1e-5, atol=1e-6)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_profiler.py b/tests/python/unittest/test_profiler.py
index 5a0baca1b684..bf745043e871 100644
--- a/tests/python/unittest/test_profiler.py
+++ b/tests/python/unittest/test_profiler.py
@@ -263,7 +263,7 @@ def check_sorting(debug_str, sort_by, ascending):
         for domain_name, domain in target_dict['Time'].items():
             lst = [item[sort_by_options[sort_by]] for item_name, item in domain.items()]
             check_ascending(lst, ascending)
-        # Memory items do not have stat 'Total' 
+        # Memory items do not have stat 'Total'
         if sort_by != 'total':
             for domain_name, domain in target_dict['Memory'].items():
                 lst = [item[sort_by_options[sort_by]] for item_name, item in domain.items()]
@@ -372,7 +372,7 @@ def check_custom_operator_profiling_multiple_custom_ops_output(debug_str):
 
 def custom_operator_profiling_multiple_custom_ops(seed, mode, file_name):
     class MyAdd(mx.operator.CustomOp):
-        def forward(self, is_train, req, in_data, out_data, aux):        
+        def forward(self, is_train, req, in_data, out_data, aux):
             self.assign(out_data[0], req[0], in_data[0] + 1)
 
         def backward(self, req, out_grad, in_data, out_data, in_grad, aux):
@@ -610,7 +610,3 @@ def test_gpu_memory_profiler_gluon():
                row['Attribute Name'] == "<unk>:":
                 assert False, "Unknown allocation entry has been encountered"
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_random.py b/tests/python/unittest/test_random.py
index efcf16dc78da..ac20a99481cf 100644
--- a/tests/python/unittest/test_random.py
+++ b/tests/python/unittest/test_random.py
@@ -22,7 +22,7 @@
 from mxnet.test_utils import verify_generator, gen_buckets_probs_with_ppf, retry, assert_almost_equal
 import numpy as np
 import random as rnd
-from common import setup_module, with_seed, random_seed, teardown
+from common import setup_module, with_seed, random_seed, teardown_module
 import scipy.stats as ss
 import unittest
 from mxnet.test_utils import *
@@ -339,7 +339,7 @@ def check_with_device(device, dtype):
            un1 = np.maximum(un1, 1e-1)
         if name == 'uniform':
            un1 = np.minimum(np.maximum(un1.reshape((un1.shape[0],un1.shape[1],-1)), p1.reshape((p1.shape[0],p1.shape[1],-1))+1e-4),
-                            p2.reshape((p2.shape[0],p2.shape[1],-1))-1e-4).reshape(un1.shape) 
+                            p2.reshape((p2.shape[0],p2.shape[1],-1))-1e-4).reshape(un1.shape)
         for use_log in [False, True]:
             test_pdf = symbol(v0, v1, is_log=use_log) if single_param else symbol(v0, v1, v2, is_log=use_log)
             forw_atol  = 1e-7 if dtype != np.float16 else 1e-3
@@ -349,7 +349,7 @@ def check_with_device(device, dtype):
             if single_param:
                 res = pdffunc(un1.reshape((un1.shape[0],un1.shape[1],-1)),
                     p1.reshape((p1.shape[0],p1.shape[1],-1))).reshape(un1.shape)
-                if use_log: 
+                if use_log:
                     res = np.log(res)
                 check_symbolic_forward(test_pdf, [un1, p1], [res], atol=forw_atol, rtol=forw_rtol, dtype=dtype)
                 if dtype == np.float64:
@@ -1035,7 +1035,3 @@ def test_sample_multinomial_num_outputs():
     assert isinstance(out, list)
     assert len(out) == 2
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_recordio.py b/tests/python/unittest/test_recordio.py
index 81561b987839..6db54d541244 100644
--- a/tests/python/unittest/test_recordio.py
+++ b/tests/python/unittest/test_recordio.py
@@ -22,7 +22,7 @@
 import tempfile
 import random
 import string
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 
 @with_seed()
 def test_recordio():
diff --git a/tests/python/unittest/test_rnn.py b/tests/python/unittest/test_rnn.py
index a5588250e515..ab333f93c918 100644
--- a/tests/python/unittest/test_rnn.py
+++ b/tests/python/unittest/test_rnn.py
@@ -308,7 +308,4 @@ def test_encode_sentences():
     print(result, vocab)
     assert vocab == {'a': 1, 'b': 2, 'c': 3, 'UNK': 0}
     assert result == [[1,2,3],[2,3,0]]
-    
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
+
diff --git a/tests/python/unittest/test_runtime.py b/tests/python/unittest/test_runtime.py
index 82e2314532b8..0bd4d4c4dcb1 100644
--- a/tests/python/unittest/test_runtime.py
+++ b/tests/python/unittest/test_runtime.py
@@ -19,14 +19,14 @@
 import sys
 from mxnet.runtime import *
 from mxnet.base import MXNetError
-from nose.tools import *
+import pytest
 
 
 def test_features():
     features = Features()
     print(features)
-    ok_('CUDA' in features)
-    ok_(len(features) >= 30)
+    assert 'CUDA' in features
+    assert len(features) >= 30
 
 
 def test_is_singleton():
@@ -39,17 +39,13 @@ def test_is_enabled():
     features = Features()
     for f in features:
         if features[f].enabled:
-            ok_(features.is_enabled(f))
+            assert features.is_enabled(f)
         else:
-            ok_(not features.is_enabled(f))
+            assert not features.is_enabled(f)
 
 
-@raises(RuntimeError)
+@pytest.mark.xfail(raises=RuntimeError)
 def test_is_enabled_not_existing():
     features = Features()
     features.is_enabled('this girl is on fire')
 
-
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_sparse_ndarray.py b/tests/python/unittest/test_sparse_ndarray.py
index 9a1fce4ff197..de06c7be0777 100644
--- a/tests/python/unittest/test_sparse_ndarray.py
+++ b/tests/python/unittest/test_sparse_ndarray.py
@@ -19,7 +19,7 @@
 
 from mxnet.ndarray import NDArray
 from mxnet.test_utils import *
-from common import setup_module, with_seed, random_seed, teardown
+from common import setup_module, with_seed, random_seed, teardown_module
 from mxnet.base import mx_real_t
 from numpy.testing import assert_allclose
 import numpy.random as rnd
@@ -1056,6 +1056,3 @@ def check_sparse_getnnz(density, axis):
         for a in axis:
             check_sparse_getnnz(d, a)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_sparse_operator.py b/tests/python/unittest/test_sparse_operator.py
index 4c4e3dbdfc51..f23f59414816 100644
--- a/tests/python/unittest/test_sparse_operator.py
+++ b/tests/python/unittest/test_sparse_operator.py
@@ -17,7 +17,7 @@
 
 from mxnet.test_utils import *
 from mxnet.base import MXNetError
-from common import setup_module, with_seed, teardown, assertRaises
+from common import setup_module, with_seed, teardown_module, assertRaises
 import random
 import warnings
 
@@ -2346,6 +2346,3 @@ def test_reshape_backward_fallback():
 
     assert_almost_equal(grad_w_nd.asnumpy(), expected_grad_nd)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_subgraph.py b/tests/python/unittest/test_subgraph.py
index 3da125a946bc..b1a7aa3f3abb 100644
--- a/tests/python/unittest/test_subgraph.py
+++ b/tests/python/unittest/test_subgraph.py
@@ -21,7 +21,7 @@
 import mxnet as mx
 import copy
 from mxnet.test_utils import *
-from common import setup_module, with_seed, teardown
+from common import setup_module, with_seed, teardown_module
 from mxnet.gluon.model_zoo.vision import get_model
 
 def make_subgraph(subg, *args):
@@ -190,6 +190,3 @@ def create_operator(self, ctx, shapes, dtypes):
     c.bind(mx.cpu(), {'a': inp}).forward()
     mx.nd.waitall()
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_subgraph_op.py b/tests/python/unittest/test_subgraph_op.py
index e414a9836ccb..bffb38e6dd5d 100644
--- a/tests/python/unittest/test_subgraph_op.py
+++ b/tests/python/unittest/test_subgraph_op.py
@@ -41,7 +41,7 @@ def network_structure_2():
     ret1 = mx.sym.cos(ret)
     ret2 = mx.sym.sin(ret)
     ret = ret1 + ret2
-    return (ret, ['data'], [(2, 3, 10, 10)]) 
+    return (ret, ['data'], [(2, 3, 10, 10)])
 
 def network_structure_3():
     # this tests whether the partitioned sym can distinguish in_args and aux_states
@@ -74,7 +74,7 @@ def network_structure_6():
     data3 = mx.sym.sin(data2)
     conv = mx.sym.Convolution(data=data1, weight=data3, kernel=(2, 2), num_filter=1)
     return (conv, ['data1'], [(3, 3, 10, 10)])
-        
+
 def network_structure_7():
     # in this graph, the subgraph node and the other two external nodes form a cycle
     data = mx.sym.Variable('data', shape=(1,))
@@ -85,7 +85,7 @@ def network_structure_7():
     ret = ret1 + ret2
     return (ret, ['data'], [(1,)])
 
-def get_graphs(): 
+def get_graphs():
     return [
             (network_structure_1(), ['Convolution']),
             (network_structure_2(), ['exp', 'sin', '_Plus', 'elemwise_add', '_plus']),
@@ -271,7 +271,7 @@ def check_subgraph_exe5(sym, subgraph_backend, op_names):
         assert_almost_equal((outputs1[i] - outputs2[i]).abs().sum().asnumpy(), np.zeros(shape=(1,)))
 
 def check_subgraph_exe6(sym, subgraph_backend, op_names):
-    """Call optimize_for to trigger graph partitioning with shapes/types, then simple_bind 
+    """Call optimize_for to trigger graph partitioning with shapes/types, then simple_bind
     and compare results of the partitioned sym and the original sym."""
     # simple_bind
     exe1 = sym.simple_bind(ctx=mx.current_context(), grad_req='null')
@@ -340,17 +340,17 @@ def check_subgraph_exe8(sym, subgraph_backend, op_names):
 
     exe2 = part_sym.bind(ctx=mx.current_context(), args=arg_array, aux_states=aux_array, grad_req='null')
     exe2.forward()
-    
+
     # compare outputs
     outputs1 = exe1.outputs
     outputs2 = exe2.outputs
     assert len(outputs1) == len(outputs2)
     for i in range(len(outputs1)):
         assert_almost_equal((outputs1[i] - outputs2[i]).abs().sum().asnumpy(), np.zeros(shape=(1,)))
-    
+
 def check_subgraph_exe9(sym, subgraph_backend, op_names):
-    """Call hybridize() to partition the graph, and then compare results of the partitioned 
-    sym and the original sym. Here do an inference before hybridizing with the subgraph_backend 
+    """Call hybridize() to partition the graph, and then compare results of the partitioned
+    sym and the original sym. Here do an inference before hybridizing with the subgraph_backend
     which means we'll pass shapes/types"""
     # create Gluon block for given symbol
     inputs = [mx.sym.var(i, dtype=mx_real_t) for i in sym[1]]
@@ -487,6 +487,3 @@ def hybrid_forward(self, F, x):
     for i in range(len(outputs1)):
         assert_almost_equal((outputs1[i] - outputs2[i]).abs().sum().asnumpy(), np.zeros(shape=(1,)))
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_symbol.py b/tests/python/unittest/test_symbol.py
index c913f5c07c9f..d5437b994ee3 100644
--- a/tests/python/unittest/test_symbol.py
+++ b/tests/python/unittest/test_symbol.py
@@ -559,6 +559,3 @@ def test_infershape_happens_for_all_ops_in_graph():
 
     assert False
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_test_utils.py b/tests/python/unittest/test_test_utils.py
index 49f0b932fdd5..98e7d94ff793 100644
--- a/tests/python/unittest/test_test_utils.py
+++ b/tests/python/unittest/test_test_utils.py
@@ -19,10 +19,11 @@
 import tempfile
 
 import mxnet as mx
-from nose.tools import *
 
+import pytest
 
-@raises(Exception)
+
+@pytest.mark.xfail(raises=Exception)
 def test_download_retries():
     mx.test_utils.download("http://doesnotexist.notfound")
 
@@ -31,4 +32,4 @@ def test_download_successful():
     tmpfile = os.path.join(tmp, 'README.md')
     mx.test_utils.download("https://raw.githubusercontent.com/apache/incubator-mxnet/master/README.md",
                            fname=tmpfile)
-    assert os.path.getsize(tmpfile) > 100
\ No newline at end of file
+    assert os.path.getsize(tmpfile) > 100
diff --git a/tests/python/unittest/test_thread_local.py b/tests/python/unittest/test_thread_local.py
index 63d97f17ebba..7e875c8fb835 100644
--- a/tests/python/unittest/test_thread_local.py
+++ b/tests/python/unittest/test_thread_local.py
@@ -221,7 +221,3 @@ def f():
     finally:
         set_np_shape(0)
 
-
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_tvm_op.py b/tests/python/unittest/test_tvm_op.py
index e325edcf4d75..55bb7cc2bd92 100644
--- a/tests/python/unittest/test_tvm_op.py
+++ b/tests/python/unittest/test_tvm_op.py
@@ -67,6 +67,3 @@ def test_tvm_broadcast_add():
             assert same(a.grad.asnumpy(), expected_grad_a)
             assert same(b.grad.asnumpy(), expected_grad_b)
 
-if __name__ == '__main__':
-    import nose
-    nose.runmodule()
diff --git a/tests/python/unittest/test_viz.py b/tests/python/unittest/test_viz.py
index 13210993014a..5c9b78a017d2 100644
--- a/tests/python/unittest/test_viz.py
+++ b/tests/python/unittest/test_viz.py
@@ -62,6 +62,3 @@ def test_plot_network():
     assert "There are multiple variables with the same name in your graph" in str(w[-1].message)
     assert "fc" in str(w[-1].message)
 
-if __name__ == "__main__":
-    import nose
-    nose.runmodule()
diff --git a/tests/requirements.txt b/tests/requirements.txt
index e16b764d2746..f7ee514760b7 100644
--- a/tests/requirements.txt
+++ b/tests/requirements.txt
@@ -1,8 +1,10 @@
 # Requirements for tests, those are installed before running on the virtualenv
 # Requirements for tests run within the qemu requirement see ci/qemu/test_requirements.txt
 mock
-nose
-nose-timer
+pytest==5.3.2
+pytest-env==0.6.2
+pytest-cov==2.8.1
+pytest-xdist==1.31.0
 ipython
 numpy
 scipy
diff --git a/tools/caffe_converter/test_converter.py b/tools/caffe_converter/test_converter.py
old mode 100755
new mode 100644
diff --git a/tools/dependencies/README.md b/tools/dependencies/README.md
index ec1e80088895..ba92cc875c8f 100644
--- a/tools/dependencies/README.md
+++ b/tools/dependencies/README.md
@@ -204,7 +204,7 @@ pip install -e python
 export NCCL_DEBUG=VERSION
 vim tests/python/gpu/test_nccl.py
 # Remove @unittest.skip("Test requires NCCL library installed and enabled during build") then run
-nosetests --verbose tests/python/gpu/test_nccl.py
+pytest --verbose tests/python/gpu/test_nccl.py
 # test_nccl.test_nccl_pushpull ... NCCL version 2.4.2+cuda10.0
 # ok
 # ----------------------------------------------------------------------
diff --git a/tools/diagnose.py b/tools/diagnose.py
old mode 100755
new mode 100644
diff --git a/tools/flakiness_checker.py b/tools/flakiness_checker.py
index 79fa3b1854f0..85eae14c9557 100644
--- a/tools/flakiness_checker.py
+++ b/tools/flakiness_checker.py
@@ -39,7 +39,7 @@ def run_test_trials(args):
 
     new_env = os.environ.copy()
     new_env["MXNET_TEST_COUNT"] = str(args.num_trials)
-    
+
     if args.seed is None:
         logging.info("No test seed provided, using random seed")
     else:
@@ -47,17 +47,17 @@ def run_test_trials(args):
 
     verbosity = "--verbosity=" + str(args.verbosity)
 
-    code = subprocess.call(["nosetests", verbosity, test_path], 
+    code = subprocess.call(["pytest", verbosity, test_path],
                            env = new_env)
-    
-    logging.info("Nosetests terminated with exit code %d", code)
+
+    logging.info("Test terminated with exit code %d", code)
 
 def find_test_path(test_file):
     """Searches for the test file and returns the path if found
     As a default, the currend working directory is the top of the search.
     If a directory was provided as part of the argument, the directory will be
     joined with cwd unless it was an absolute path, in which case, the
-    absolute path will be used instead. 
+    absolute path will be used instead.
     """
     test_file += ".py"
     test_path = os.path.split(test_file)
@@ -66,7 +66,7 @@ def find_test_path(test_file):
     for (path, dirs, files) in os.walk(top):
         if test_path[1] in files:
             return  os.path.join(path, test_path[1])
-    raise FileNotFoundError("Could not find " + test_path[1] + 
+    raise FileNotFoundError("Could not find " + test_path[1] +
                             "in directory: " + top)
 
 class NameAction(argparse.Action):
@@ -82,12 +82,12 @@ def __call__(self, parser, namespace, values, option_string=None):
 
 def parse_args():
     parser = argparse.ArgumentParser(description="Check test for flakiness")
-    
+
     parser.add_argument("test", action=NameAction,
                         help="file name and and function name of test, "
                         "provided in the format: <file-name>.<test-name> "
                         "or <directory>/<file>:<test-name>")
-    
+
     parser.add_argument("-n", "--num-trials", metavar="N",
                         default=DEFAULT_NUM_TRIALS, type=int,
                         help="number of test trials, passed as "
@@ -95,11 +95,11 @@ def parse_args():
 
     parser.add_argument("-s", "--seed", type=int,
                         help="test seed, passed as MXNET_TEST_SEED, "
-                        "defaults to random seed") 
+                        "defaults to random seed")
 
     parser.add_argument("-v", "--verbosity",
                         default=DEFAULT_VERBOSITY, type=int,
-                        help="logging level, passed to nosetests")
+                        help="logging level, passed to pytest")
 
 
     args = parser.parse_args()
diff --git a/tools/im2rec.py b/tools/im2rec.py
old mode 100755
new mode 100644
diff --git a/tools/ipynb2md.py b/tools/ipynb2md.py
old mode 100755
new mode 100644
diff --git a/tools/launch.py b/tools/launch.py
old mode 100755
new mode 100644
diff --git a/tools/parse_log.py b/tools/parse_log.py
old mode 100755
new mode 100644