* [HEADER] ASF header dir=include

* [HEADER] ASF Header dir=src

* [HEADER] ASF Header -dir=python

* [HEADER] ASF header dir=topi

* [HEADER] ASF Header dir=nnvm

* [HEADER] ASF Header -dir=tutorials

* [HEADER] ASF Header dir=tests

* [HEADER] ASF Header -dir=docker

* fix whitespace

* [HEADER] ASF Header -dir=jvm

* [HEADER] ASF Header -dir=web

* [HEADER] ASF Header --dir=apps

* [HEADER] ASF Header --dir=vta

* [HEADER] ASF Header -dir=go

* temp

* [HEADER] ASF Header --dir=rust

* [HEADER] Add ASF Header --dir=cmake

* [HEADER] ASF Header --dir=docs

* [HEADER] Header for Jenkinsfile

* [HEADER] ASF Header to toml and md

* [HEADER] ASF Header to gradle

* Finalize rat cleanup

* Fix permission

* Fix java test

* temporary remove nnvm onnx test

* [NNVM][TEST] Numerical gradient testing

* [NNVM][TEST] Make some tests a little faster

* Fix the failing test_top_level3

* Target exclusion for the check_function

* Try to ignore singularities

* grad_input_vars now can't contain shapes

* Don't pass unnecessary grad_input_vars to check_function

* Multiple outputs; fixes; testing of check_function

* Use numerical_grads_params to pass parameters to numgrad checker

* Fail when no action is requested excplicitly

* Pass additional params to functions

* Silence the linter issue

* Simplified numgrad checking

* Improved docs for check_function

* Fixed the error message when no dtype is provided

* Several fixes

* Tests with shape/dtype inference for inputs

* Don't check dense's grads on cuda

* Raise an error if output dtypes haven't been inferred

* Moved shape/dtype inference into a separate function; use float32 as fallback

* Remove redundant dtype=float32

* Fix multiple outputs

* Use check_function in the rest of the test_top_level1

[NNVM] Fix softmax gradient

* Update docs

* Add backward operator to enable backward graph

* Fix testing

* Refactor top level1 test code

* Fix format

* Test

* Added zeros ones op

* Register fill_like operator

* Fix unit test

* vgg16 success

* remove six.PY2, use sys.version_info;
convert_activation() accepts activation type name(str, e.g. 'relu') as input;

* add convert_merge

* fix convert_batchnorm;
improve tests

* fix lint

* add numpy-style pad operator

* deal with asymmetry padding

* resnet50 success

* fix pool_convert; xception passes test

* update tvm

* fix bias error; all tests pass

* use > >, not >>

* add transform

* fix

* update doc

* Update tvm

* register pool, global_pool; add mobilenet example

* tests of pool and global_pool

* use new API of runtime module

* small fix

* [TOP] Add dense, batchnorm

* update tvm

* [RUNTIME][COMPILER] Formal compiler pipeline, runtime wrapper module

* more detailed comments

* register depthconv, elemwise

* use global elemwise schedule for relu