* allow amp and recompute working together

API/OP (center_loss, fluid.one_hot, prroi_pool, roi_pool, ctc_greed_decoder) error message enhancement (#23794)

* error message enchanced, test=develop

* error message enchanced for APIs, test=develop

* error message enhanced for roi_pool, test=develop

* update added code, test=develop

* update fluid/input.py, test=develop

* update fluid/input.py ,test=develop

* fix code style, test=develop

* fix lower converage, test=develop

* error message enhanced for roi pool, test=develop

Correct the name [`FLAGS_sync_nccl_allreduce`](https://www.paddlepaddle.org.cn/documentation/docs/zh/advanced_guide/flags/others_cn.html#flags-sync-nccl-allreduce) based on the information from our official website.

As the title.

* fix bug in reduce op test=develop

* fix bug in reduce op test=develop

* API(fused_embedding_fc_lstm) error message enhancement. test=develop

C++ API enhancement.

* Refine code, test=develop

* Refine code. test=develop

Enhance print op c++ error message.

* error message enhancement create_parameter create_global_var, test=develop

* use six.PY2 to use long, test=develop

* fix input dtype for create_parameter, test=develop

Fix the compile problem in norm op in the windows env, add delete some invalid argument 

* refine error message, test=develop

* fix judge pslib & ranspiler

* enhance the error message of box_coder, test=develop

* enhance error message in box_decoder_and_assign

* enhance error message of collect_fpn_proposals

* enhance errror message of distribute_fpn_proposals

* enhance error message of multiclass_nms, test=develop

* add nn.interpolate, test=develop

* fix interpolate typo, test=develop

* formate code, test=develop

* fix unitest, test=develop

* add test layers, test=develop

* add test layers, test=develop

* extract common function, test=develop

* reduce the threads for cuda10, test=develop

* update unitest, test=develop

* polish unitest, test=develop

* add dygraph test, test=develop

* format description, test=develop

* add 5D input check, test=develop

* fix doc, test=develop

* enrich count enforce scripts, test=develop, test=document_fix

* add usage & result, test=develop, test=document_fix

test=develop

* enhance error message test=develop

* enhance error message test=develop

* change to INOUT_CHECK  test=develop

* add cudnn compatiable rnn cell api for dygraph

* update sample code

* update some typos

* fix specify name in param_attr problem

* add pre-commit check

* remove duplicate import, test=develop

* add unittest coverage, test=develop

* make code more tight, test=develop

* cudnn_compatibale -> use_cudnn_impl, test=develop

* change api name, test=develop

* test=develop fix api doc error

Fix the scale fp16 dtype support for scale op, and delete the test case for fp16 case

* `l1_norm` error message enhancement

test=develop

* `norm` error message enhancement

test=develop

* `squared_l2_norm` error message enhancement

test=develop

* `squared_l2_distance` error message enhancement

test=develop

* `conv_shift_op` error message enhancement

test=develop

* `sample_logits_op` error message enhancement

test=develop

* Fix format

test=develop

* use `PADDLE_ENFORCE_EQ` for '||' conditional

test=develop

* Fix string format

test=develop

* Add test for `sample_logits`

test=develop

* Add error message

test=develop

* Use ordered dict for outputs

test=develop

* Hard code fetch list so that order is deterministic

test=develop

* Fix license header year

test=develop

eagerly release cuda resources before cuda enviroment destroying
test=develop

* enhance attention_lstm and param_attr error message
* fix: fix param_attr type check

* add approximation for gelu, test=develop

* add functional conv

* add test and doc for function convs, test=develop

* update ConvTransposeOp's InferShape and error message, test=develop

* add hsigmoid, row_conv in paddle.nn(functional and layer), test=develop

* fix hyperlinks in docstring

* enhance error message, test=develop

fix sample code diff for OP(retinanet_detection_output, retinanet_target_assign), fix `gamma` default value for OP(sigmoid_focal_loss) (#23744)

* fix sample code diff for OP(retinanet_detection_output, retinanet_target_assign), fix  default value for OP(sigmoid_focal_loss). test=develop

* fix float value. test=develop

* add int type for gamma and alpha.test=develop

* optimize compare and logical ops error info
* add out and cond dtype test

* correct reader device index, test=develop

* fix async executor scope var initialization, test=develop

* add hard_sigmoid trt op converter

* add hard_swish op converter and plugin. test=develop

* add macro to adapt lower trt version. test=develop

* Add variance computing layer for tensor

* Format input data types

* Remove unnecessary expand operation

* Some fixes in doc