avoid the default one that usually overwrites manually set ones

test=develop

* add detect and end2end code

* update the scale for coodinates restore

* fix merge bug with dev.

* fix merge bug with dev.

* test=develop

* fix code style test=develop

* fix code style test=develop

* test=develop

* test=develop

* test=develop

* Add affine channel op.
* Update code and add Python API.
test=develop
* Update API.spec
test=develop

Add rpc timeline

* fill constant add infervarshape, lookuptable clone lr var

* test=develop

* add lookuptable ut, test=develop

* bug fix in transpliler about async with lookup table

* test=develop

test=develop

test=develop

test=develop

This can cause a lot of problem:
1. Wrong operator implementation, Op can get a wrong type without failure.
2. Anytype can be Get without defined in VarType.

Also fix wrong STEP_SCOPE usage.

test=develop

test=develop

* set the right shape of selected_rows
test=develop

* enhance check

* fix activation_op

* remove cast

* use ShareDimInfo replace SetDim and ShareLod

* use ShareDimAndLod
test=develop

* follow comment

test=develop

* check whether the input has lod
test=develop

* Split ShareDimAndLod

test=develop

* checkout clip.py
test=develop

test=develop

* Fix bug in reduce_op caused by PR #13534
* Fix output shape and enhance unit test.
test=develop

* "avoid copy back to cpu"

* "add infinity support"

* "fix ci"

* "add cpu macro"

* rerun ci; test=develop

* "fix api"
test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* test=develop

- Work in progress on unit test for fused_embedding_fc_lstm op

- Added bias caching and ref x computing

- Small update unit test

- temporary storage

- Fix to batchcompute

- Cosmetic fixes

- Style fixes

* parallel run dist ci

* test=develop

* update test=develop

This reverts commit 21bb9e91, reversing
changes made to 3fa68dc1.

test=develop

* refine reduce by cub
* optimize KernelDepthwiseConvFilterGrad
* optimize depthwise conv and reduce mean and reduce sum
* fix bug: dilation
* cuda arch and cuda 8 compatible

test=develop

test=develop

test=develop