bukejiyu

[paddle Inference]add_group_norm_pass

1

### PR Category Inference ### PR Types Others ### Description 1.增加add_group_norm_silu kernel 支持传入残差和激活 2.增加add_group_norm pattern 3.增加对应单测 pcard-71500

[Paddle inference]conv2d_add_act_pass support cutlass

1

### PR Category Inference ### PR Types Others ### Description pcard-71500 1.pass支持调入cutlass kernel

【xft_test】add xft_test

1

### PR types Others ### PR changes Others ### Description

gcc8.2 编译报错

1


[cpu]update xft doc

3

### PR types Others ### PR changes Others ### Description 增加xft使用readme

[paddle inference]add flash_attn_2 custom kernel

4

### PR types Others ### PR changes Others ### Description 增加flash_attn_2 自定义算子

add mqa and flash_2

3

### PR types Others ### PR changes Others ### Description add flash2 和 mqa

add append_attn ut

3

#### Before submitting - [ ] Lint code. If there are lint issues, please format the code first. ```shell # Install and register `pre-commit` in the project folder pip install...

update doc

2

#### Before submitting - [ ] Lint code. If there are lint issues, please format the code first. ```shell # Install and register `pre-commit` in the project folder pip install...