[DOC] Fix warnings in tutorials and turn on -W (#19624)

ci/docker/runtime_functions.sh

@@ -712,6 +712,7 @@ sanity_cpp() {
 
 sanity_python() {
     set -ex
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     python3 -m pylint --rcfile=ci/other/pylintrc --ignore-patterns=".*\.so$$,.*\.dll$$,.*\.dylib$$" python/mxnet
     OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -n 4 tests/tutorials/test_sanity_tutorials.py
 }
@@ -728,7 +729,7 @@ cd_unittest_ubuntu() {
     export MXNET_SUBGRAPH_VERBOSE=0
     export MXNET_ENABLE_CYTHON=0
     export CD_JOB=1 # signal this is a CD run so any unecessary tests can be skipped
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
 
     local mxnet_variant=${1:?"This function requires a mxnet variant as the first argument"}
 
@@ -767,7 +768,7 @@ unittest_ubuntu_python3_cpu() {
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
     export MXNET_ENABLE_CYTHON=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
     MXNET_ENGINE_TYPE=NaiveEngine \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --cov-append --verbose tests/python/unittest
@@ -781,7 +782,7 @@ unittest_ubuntu_python3_cpu_mkldnn() {
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
     export MXNET_ENABLE_CYTHON=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
     MXNET_ENGINE_TYPE=NaiveEngine \
                      OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --cov-append --verbose tests/python/unittest
@@ -797,7 +798,7 @@ unittest_ubuntu_python3_gpu() {
     export MXNET_SUBGRAPH_VERBOSE=0
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
     export MXNET_ENABLE_CYTHON=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator and not test_amp_init.py' -n 4 --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
@@ -816,7 +817,7 @@ unittest_ubuntu_python3_gpu_cython() {
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
     export MXNET_ENABLE_CYTHON=1
     export MXNET_ENFORCE_CYTHON=1
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     check_cython
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator and not test_amp_init.py' -n 4 --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
@@ -834,7 +835,7 @@ unittest_ubuntu_python3_gpu_nocudnn() {
     export MXNET_SUBGRAPH_VERBOSE=0
     export CUDNN_OFF_TEST_ONLY=true
     export MXNET_ENABLE_CYTHON=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator and not test_amp_init.py' -n 4 --durations=50 --cov-report xml:tests_gpu.xml --verbose tests/python/gpu
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
@@ -846,13 +847,15 @@ unittest_ubuntu_python3_gpu_nocudnn() {
 
 unittest_cpp() {
     set -ex
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     build/tests/mxnet_unit_tests
 }
 
 unittest_centos7_cpu() {
     set -ex
     source /opt/rh/rh-python36/enable
     cd /work/mxnet
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     OMP_NUM_THREADS=$(expr $(nproc) / 4) python -m pytest -m 'not serial' -k 'not test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --verbose tests/python/unittest
     MXNET_ENGINE_TYPE=NaiveEngine \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) python -m pytest -m 'not serial' -k 'test_operator' -n 4 --durations=50 --cov-report xml:tests_unittest.xml --cov-append --verbose tests/python/unittest
@@ -865,7 +868,7 @@ unittest_centos7_gpu() {
     source /opt/rh/rh-python36/enable
     cd /work/mxnet
     export CUDNN_VERSION=${CUDNN_VERSION:-7.0.3}
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
         OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -m 'not serial' -k 'not test_operator and not test_amp_init.py' -n 4 --durations=50 --cov-report xml:tests_gpu.xml --cov-append --verbose tests/python/gpu
     MXNET_GPU_MEM_POOL_TYPE=Unpooled \
@@ -879,7 +882,7 @@ integrationtest_ubuntu_cpu_onnx() {
 	set -ex
 	export PYTHONPATH=./python/
 	export MXNET_SUBGRAPH_VERBOSE=0
-	export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
 	python3 tests/python/unittest/onnx/backend_test.py
 	OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -n 4 tests/python/unittest/onnx/mxnet_export_test.py
 	OMP_NUM_THREADS=$(expr $(nproc) / 4) pytest -n 4 tests/python/unittest/onnx/test_models.py
@@ -894,7 +897,7 @@ integrationtest_ubuntu_cpu_dist_kvstore() {
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
     export MXNET_USE_OPERATOR_TUNING=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     cd tests/nightly/
     python3 ../../tools/launch.py -n 7 --launcher local python3 dist_sync_kvstore.py --type=gluon_step_cpu
     python3 ../../tools/launch.py -n 7 --launcher local python3 dist_sync_kvstore.py --type=gluon_sparse_step_cpu
@@ -910,6 +913,7 @@ integrationtest_ubuntu_cpu_dist_kvstore() {
 
 integrationtest_ubuntu_gpu_dist_kvstore() {
     set -ex
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     pushd .
     cd /work/mxnet/python
     pip3 install -e .
@@ -929,7 +933,7 @@ integrationtest_ubuntu_gpu_byteps() {
     git clone -b v0.2.3 https://github.com/bytedance/byteps ~/byteps
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     cd tests/nightly/
 
     export NVIDIA_VISIBLE_DEVICES=0
@@ -952,7 +956,7 @@ test_ubuntu_cpu_python3() {
     set -ex
     pushd .
     export MXNET_LIBRARY_PATH=/work/build/libmxnet.so
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     VENV=mxnet_py3_venv
     virtualenv -p `which python3` $VENV
     source $VENV/bin/activate
@@ -976,7 +980,7 @@ unittest_ubuntu_python3_arm() {
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
     export MXNET_ENABLE_CYTHON=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     python3 -m pytest -n 2 --verbose tests/python/unittest/test_engine.py
 }
 
@@ -1010,31 +1014,31 @@ test_rat_check() {
 nightly_test_KVStore_singleNode() {
     set -ex
     export PYTHONPATH=./python/
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     tests/nightly/test_kvstore.py
 }
 
 #Test Large Tensor Size
 nightly_test_large_tensor() {
     set -ex
     export PYTHONPATH=./python/
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     pytest --timeout=0 --forked tests/nightly/test_np_large_array.py
 }
 
 #Tests Model backwards compatibility on MXNet
 nightly_model_backwards_compat_test() {
     set -ex
     export PYTHONPATH=/work/mxnet/python/
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     ./tests/nightly/model_backwards_compatibility_check/model_backward_compat_checker.sh
 }
 
 #Backfills S3 bucket with models trained on earlier versions of mxnet
 nightly_model_backwards_compat_train() {
     set -ex
     export PYTHONPATH=./python/
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     ./tests/nightly/model_backwards_compatibility_check/train_mxnet_legacy_models.sh
 }
 
@@ -1043,7 +1047,7 @@ nightly_tutorial_test_ubuntu_python3_gpu() {
     cd /work/mxnet/docs
     export BUILD_VER=tutorial
     export MXNET_DOCS_BUILD_MXNET=0
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     make html
     export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
     export MXNET_SUBGRAPH_VERBOSE=0
@@ -1055,7 +1059,7 @@ nightly_tutorial_test_ubuntu_python3_gpu() {
 
 nightly_estimator() {
     set -ex
-    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    export DMLC_LOG_STACK_TRACE_DEPTH=100
     cd /work/mxnet/tests/nightly/estimator
     export PYTHONPATH=/work/mxnet/python/
     pytest test_estimator_cnn.py

docs/python_docs/python/Makefile_sphinx

@@ -37,7 +37,7 @@ endif # $(NUMJOBS)
 # End number of processors detection
 
 # You can set these variables from the command line.
-SPHINXOPTS    = -j$(NPROCS) -c ../scripts --keep-going
+SPHINXOPTS    = -j$(NPROCS) -c ../scripts --keep-going -W
 SPHINXBUILD   = sphinx-build
 PAPER         =
 BUILDDIR      = _build

docs/python_docs/python/api/gluon/rnn/index.rst

@@ -25,7 +25,6 @@ Build-in recurrent neural network layers are provided in the following two modul
     :nosignatures:
 
     mxnet.gluon.rnn
-    mxnet.gluon.contrib.rnn
 
 .. currentmodule:: mxnet.gluon
 

docs/python_docs/python/api/kvstore/index.rst

@@ -15,9 +15,34 @@
    specific language governing permissions and limitations
    under the License.
 
-mxnet.kvstore
-=============
+KVStore: Communication for Distributed Training
+===============================================
+.. currentmodule:: mxnet.kvstore
 
-.. automodule:: mxnet.kvstore
-    :members:
-    :autosummary:
+
+Horovod
+=======
+
+.. autosummary::
+   :toctree: generated/
+
+   Horovod
+
+BytePS
+======
+
+.. autosummary::
+   :toctree: generated/
+
+   BytePS
+
+
+KVStore Interface
+=================
+
+.. autosummary::
+   :toctree: generated/
+
+   KVStore
+   KVStoreBase
+   KVStoreServer

docs/python_docs/python/api/npx/index.rst

@@ -84,6 +84,7 @@ More operators
    sigmoid
    smooth_l1
    softmax
+   log_softmax
    topk
    waitall
    load

docs/python_docs/python/tutorials/deploy/export/onnx.md

@@ -28,7 +28,7 @@ In this tutorial, we will learn how to use MXNet to ONNX exporter on pre-trained
 ## Prerequisites
 
 To run the tutorial you will need to have installed the following python modules:
-- [MXNet >= 1.3.0](/get_started)
+- [MXNet >= 1.3.0](https://mxnet.apache.org/get_started)
 - [onnx]( https://github.com/onnx/onnx#installation) v1.2.1 (follow the install guide)
 
 *Note:* MXNet-ONNX importer and exporter follows version 7 of ONNX operator set which comes with ONNX v1.2.1.
@@ -44,7 +44,7 @@ logging.basicConfig(level=logging.INFO)
 
 ## Downloading a model from the MXNet model zoo
 
-We download the pre-trained ResNet-18 [ImageNet](http://www.image-net.org/) model from the [MXNet Model Zoo](/api/python/docs/api/gluon/model_zoo/index.html).
+We download the pre-trained ResNet-18 [ImageNet](http://www.image-net.org/) model from the [MXNet Model Zoo](../../../api/gluon/model_zoo/index.rst).
 We will also download synset file to match labels.
 
 ```{.python .input}
@@ -59,7 +59,7 @@ Now, we have downloaded ResNet-18 symbol, params and synset file on the disk.
 
 ## MXNet to ONNX exporter API
 
-Let us describe the MXNet's `export_model` API. 
+Let us describe the MXNet's `export_model` API.
 
 ```{.python .input}
 help(onnx_mxnet.export_model)
@@ -74,7 +74,7 @@ export_model(sym, params, input_shape, input_type=<type 'numpy.float32'>, onnx_f
     Exports the MXNet model file, passed as a parameter, into ONNX model.
     Accepts both symbol,parameter objects as well as json and params filepaths as input.
     Operator support and coverage - https://cwiki.apache.org/confluence/display/MXNET/MXNet-ONNX+Integration
-    
+
     Parameters
     ----------
     sym : str or symbol object
@@ -89,7 +89,7 @@ export_model(sym, params, input_shape, input_type=<type 'numpy.float32'>, onnx_f
         Path where to save the generated onnx file
     verbose : Boolean
         If true will print logs of the model conversion
-    
+
     Returns
     -------
     onnx_file_path : str
@@ -145,6 +145,6 @@ model_proto = onnx.load_model(converted_model_path)
 checker.check_graph(model_proto.graph)
 ```
 
-If the converted protobuf format doesn't qualify to ONNX proto specifications, the checker will throw errors, but in this case it successfully passes. 
+If the converted protobuf format doesn't qualify to ONNX proto specifications, the checker will throw errors, but in this case it successfully passes.
 
 This method confirms exported model protobuf is valid. Now, the model is ready to be imported in other frameworks for inference!

docs/python_docs/python/tutorials/getting-started/crash-course/0-introduction.md

@@ -33,9 +33,9 @@ Apache MXNet is an open-source deep learning framework that provides a comprehen
 
 #### Tensors A.K.A Arrays
 
-Tensors give us a generic way of describing $n$-dimensional **arrays** with an arbitrary number of axes. Vectors, for example, are first-order tensors, and matrices are second-order tensors. Tensors with more than two orders(axes) do not have special mathematical names. The [ndarray](https://mxnet.apache.org/versions/1.7/api/python/docs/api/ndarray/index.html) package in MXNet provides a tensor implementation. This class is similar to NumPy's ndarray with additional features. First, MXNet’s `NDArray` supports fast execution on a wide range of hardware configurations, including CPU, GPU, and multi-GPU machines where as NumPy only supports CPU computation. Second, MXNet’s `NDArray` executes code lazily, allowing it to automatically parallelize multiple operations across the available hardware.
+Tensors give us a generic way of describing $n$-dimensional **arrays** with an arbitrary number of axes. Vectors, for example, are first-order tensors, and matrices are second-order tensors. Tensors with more than two orders(axes) do not have special mathematical names. The [NP](../../../api/np/index.rst) package in MXNet provides a NumPy-compatible tensor implementation, `np.ndarray` with additional features. First, MXNet’s `np.ndarray` supports fast execution on a wide range of hardware configurations, including CPU, GPU, and multi-GPU machines where as NumPy only supports CPU computation. Second, MXNet’s `np.ndarray` executes code lazily, allowing it to automatically parallelize multiple operations across the available hardware.
 
-You will get familiar to arrays in the [next section](1-nparray.md) of this crash course.
+You will get familiar to arrays in the [next section](./1-nparray.ipynb) of this crash course.
 
 ### Computing paradigms
 
@@ -44,9 +44,9 @@ You will get familiar to arrays in the [next section](1-nparray.md) of this cras
 Neural network designs like [ResNet-152](https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/He_Deep_Residual_Learning_CVPR_2016_paper.pdf) have a fair degree of regularity. They consist of _blocks_ of repeated (or at least similarly designed) layers; these blocks then form the basis of more complex network designs. A block can be a single layer, a component consisting of multiple layers, or the entire complex neural network itself! One benefit of working with the block abstraction is that you can combine blocks into larger artifacts; often recursively. By defining code to generate blocks of arbitrary complexity on demand, you can write surprisingly compact code and still implement complex neural networks.
 
 
-From a programming standpoint, a block is represented by a class and [Block](https://mxnet.apache.org/versions/1.7/api/python/docs/api/gluon/nn/index.html#mxnet.gluon.nn.Block)  is the base class for all neural networks layers in MXNet. Any subclass of it must define a forward propagation function that transforms its input into output and must store any necessary parameters if required.
+From a programming standpoint, a block is represented by a class and [Block](../../../api/gluon/block.rst#mxnet.gluon.Block)  is the base class for all neural networks layers in MXNet. Any subclass of it must define a forward propagation function that transforms its input into output and must store any necessary parameters if required.
 
-You will see more about blocks in [Array](1-nparray.md) and [Create neural network](2-create-nn.md) sections.
+You will see more about blocks in [Array](./1-nparray.ipynb) and [Create neural network](./2-create-nn.ipynb) sections.
 
 #### HybridBlock
 
@@ -62,7 +62,7 @@ You can learn more about the difference between symbolic vs. imperative programm
 
 When designing MXNet, developers considered whether it was possible to harness the benefits of both imperative and symbolic programming. The developers believed that users should be able to develop and debug using pure imperative programming, while having the ability to convert most programs into symbolic programming to be run when product-level computing performance and deployment are required. 
 
-In hybrid programming, you can build models using either the [HybridBlock](https://mxnet.apache.org/versions/1.7/api/python/docs/api/gluon/hybrid_block.html) or the [HybridSequential](https://mxnet.apache.org/versions/1.6/api/python/docs/api/gluon/nn/index.html#mxnet.gluon.nn.HybridSequential) and [HybridConcurrent](https://mxnet.incubator.apache.org/versions/1.7/api/python/docs/api/gluon/contrib/index.html#mxnet.gluon.contrib.nn.HybridConcurrent) classes. By default, they are executed in the same way Block or Sequential  and Concurrent  classes are executed in imperative programming. When the  `hybridize`  function is called, Gluon will convert the program’s execution into the style used in symbolic programming. This allows one to optimize the compute-intensive components without sacrifices in the way a model is implemented. In fact, most models can make use of hybrid programming’s execution style.
+In hybrid programming, you can build models using either the [HybridBlock](../../../api/gluon/hybrid_block.rst#mxnet.gluon.HybridBlock) or the [HybridSequential](../../../api/gluon/nn/index.rst#mxnet.gluon.nn.HybridSequential) and [HybridConcatenate](../../../api/gluon/nn/index.rst#mxnet.gluon.nn.HybridConcatenate) classes. By default, they are executed in the same way [Block](../../../api/gluon/block.rst#mxnet.gluon.Block) or [Sequential](../../../api/gluon/nn/index.rst#mxnet.gluon.nn.Sequential) and [Concurrent](../../../api/gluon/nn/index.rst#mxnet.gluon.nn.Concatenate) classes are executed in imperative programming. When the  `hybridize`  function is called, Gluon will convert the program’s execution into the style used in symbolic programming. This allows one to optimize the compute-intensive components without sacrifices in the way a model is implemented. In fact, most models can make use of hybrid programming’s execution style.
 
 You will learn more about hybrid blocks and use them in the upcoming sections of the course.
 
@@ -72,7 +72,7 @@ Gluon is an imperative high-level front end API in MXNet for deep learning that
 
 ## Next steps
 
-Dive deeper on [array representations](1-nparray.md) in MXNet.
+Dive deeper on [array representations](./1-nparray.ipynb) in MXNet.
 
 ## References
 1.  [Dive into Deep Learning](http://d2l.ai/)