* MKLDNN pool2d OP kernel added

* conv2d and pool2d MKLDNN kernels renamed

* MKLDNN conv2d kernel refactoring

* fix mac build error

* override GetExpectedKernelType for cast op

* fix typo

* add cuda unittest

Fix sparse update memory error for distributed training

* MKLDNN conv2 OP kernel added

* TODOs added

* mkldnn conv2d OP refactor

* CanCUDNNBeUsed and CanMKLDNNBeUsed moved

* Fix mAP evaluator bug.

* Fix bug in detection mAP evaluator.

* Fix unit testing.

* Support to set background label index in detection mAP op.

* change learning_rate_decay to learning_rate_scheduler
* optimize code
* change nn.cast to tensor.cast

Implement detection mAP evaluator wrapper and unify label format between SSD loss and mAP evaluator  (#8736)

* Implement mAP evalutor Python interface.

* Fix unit testing and uniy label format between SSD loss and mAP evalutor.

* Update doc.

* fix ssd problems

* fix box decoder op

* fix dimension problem in detection tests

* update detection doc

* Update detection doc

* Update detection doc

* update detection doc

* update detection doc

When the second argument contains batch dimension, the axis should be 0.

Also makes elementwise ops more tolerant at handling tensors with trailing
singular dimensions.

* Enhance bipartite_match_op to support argmax matching after bipartite matching.

* Fix typo error.