GitLab

- Implement core modules: Linear, Embedding, RMSNorm
- Add PyTorch-like macros for module and parameter definition
  - INFINICORE_NN_MODULE for single module declaration
  - INFINICORE_NN_MODULE_VEC for module vectors
  - INFINICORE_NN_PARAMETER for parameter declaration
  - Corresponding INIT macros for initialization
- Implement hierarchical module system with dynamic path generation
- Add state_dict() and load_state_dict() support
- Refactor module design: protected registration methods, removed path_ member
- Add comprehensive test suite including TinyLlama integration
- All parameters are protected with public accessors

Co-authored-by: zhuyue <zhuyue@qiyuanlab.com>

* issue/472: p800 ccl

* issue/472: 删掉无用操作

* issue/472: fix format

* issue/472: memcpy h2h case

* 适配p800软件栈

* 支持p800上编译手写算子，重构elementwise 算子组件

* fix format

* delete include xtdk_io

* 支持 elementwise float16 bfloat16 数据类型

* 修改自定义 size_t 和 ptrdiff_t 中基础数据类型

* p800 拆分 kernel

Issue/277: 为 ReLU 的九齿实现接入天数设备

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: PanZezhong <panzezhong@qiyuanlab.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>

Signed-off-by: YdrMaster <ydrml@hotmail.com>