* add maxunpool3d op

* update doc for maxunpool3d op

* update doc for maxunpool3d op

* update doc for maxunpool3d op

* update sample code for maxunpool3d

* add maxunpool1d op

* update some code for maxunpool1d

* add mish operator and api

* remove redundant code and modify grad_atol of mish unittest

* modify mish code to be consistent with other activation implementation

* add fold opereators, test=develop

* add fold opereators, test=develop

* add fold opereators, test=develop

* update fold op error test, test=develop

* fix unitext, test=develop

* fix unitext, test=develop

Add key_padding_mask and attn_mask in sparse_attention Api

1.Key padding mask is a tensor with dimensions [batch_size, seq_len], and attention mask is a tensor with dimensions [seq_len, seq_len]. The data types of the two masks are consistent with Q, K, and V, which are float32 or float64. If the value in Mask is 0, it means that the position needs to be masked.

2.The changed files are mainly paddle/fluid/operators/sparse_attention_op.cu and python/paddle/fluid/tests/unittests/test_sparse_attention_op.py. sparse_attention has three parts: sddmm, softmax, and dsd. Adding the mask operation only needs to modify the softmax. It has no effect on the other two parts. In addition, in order to test the mask function, related tests has been added.

* add hinge_embedding_loss

* fix test_API

* test_API succeed

* add English doc

* fixed using of expired fluid api

* fix doc

* fix doc and rm python/paddle/fluid/layers/loss.py

* get raw python/paddle/fluid/layers/loss.py back

* fix Examples bug in English doc

* unique -> flatten

* fix api code

* fix English doc

* fix functional loss English doc

* fix Example doc

* .numpy() -> paddle.unique()

* fix unique

* fix label_item_set

* modified judgment equation

* Got a beautiful loss equation

* use paddle.to_tensor

* fix loss and add static check

* fix loss and add static check

* delta -> margin

* support data_format='NHWC' for prelu channel mode

* fix dropout static when axis != None

* update dropout test

* add dropout test

* fix test

* Update test_dropout_op.py

* Update test_dropout_op.py

* fix testcase

* fix testcase

* Update test_dropout_op.py

* fix testcase

* fix testcase

* optimize perf

* add new test

* fix testcase

* add zeropad2d v0.1

* add zeropad2d v0.2

* add zeropad2d v0.3

* add zeropad2d v0.3

* add zeropad2d v0.3

* add zeropad2d v0.4

* add zeropad2d v0.5

* add zeropad2d v0.5 codestyle

* add zeropad2d v0.5 codestyle

* add zeropad2d v0.6 functional

* add zeropad2d v0.6 functional

* add zeropad2d v0.6 functional

* Fix the slow running speed of kl_div when option reduction is set

* fix unittest coverage

* modify sparse_attention docs, test=develop

* add warning

* add warning ,test=document_fix

* fix bug:
1. atten: set the default value of attn_dropout_rate to None
2. ffn: add activation parameter

* Revert "Align CTC grad scale same with ESPNet (#34729)"

This reverts commit 10f9644c.

* ctc grad compute on gpu

将 #35905 和 #35843 PR中新增的的python api接口移到incubate目录下。

这个PR只包含fused_feedforward前向的代码。

相关kernel实现：fused_dropout_act_bias, fused_residual_dropout_bias, fused_layernorm_residual_dropout_bias

fused_feedforward是一个融合算子，该算子对transformer模型的feed forward层的算子进行融合和封装，使得前端只呈现一个接口，通过融合减少部分访存和kernel launch的时间，以此提升性能。

功能：本PR的目标是提高attention模块的计算性能。
为了减少框架层对op的调度开销，本PR通过在C++层手动实现attention模块，对外提供attention 大op；
为了减少防存开销，本PR采取了两种优化方法：
（1）在q,k,v计算时通过共享输入X，将该处的gemm，transpose和bias add从三次调用减少为一次；
（2）使用kernel融合优化技术，在不同cuda kernel之间通过寄存器传输数据；

* update

* update

* update

* try make CI pass

* doc typo

* update doc string

Add paddle.nn.functional.sparse_attention API

    本个PR主要将sparse_attention功能在python层进行了一层封装，OP的主体代码见：#PR35676

    此外，对于封装的python 接口，增加了相应的单测。

* fix pad tuple

* fix format

* fix unpool doc, test=document_fix

* fix typo for python example, test=document_fix

* [NPU] fix depthwise_conv2d_grad, test=develop

* remove debug files, test=develop