GitLab

issue/716: Add save feature for existing test cases

issue/746: 修复causal_softmax在长宽在1024边缘的计算错误

Signed-off-by: Ceng23333 <441651826@qq.com>

* issue/739 - support batched RoPE on Nvidia and CPU

* issue/739 - metax, moore batched rope

* issue/739 - adjust metax flags

* issue/739 - added a rope module interface to forward inplace in output tensor

issue/741  暴露 infinicore_cpp_api 类继承

issue/722 - adjusted cuda rearrange for shape (8, 4, 20, 64)

issue/713 - 为c++添加 RowParallelLinear 和 ColumnParallelLinear

issue/730 - optimized debug

 Module 支持单个张量加载

Issue/714 - rearrange和random_sample测试修改

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

issue/704 - Add ccl supprt for maca with mc api

issue/700  算子执行时根据张量 set device， CPU时无操作

fix: correct the typo for the explanation of the option `--iluvatar-gpu`

Signed-off-by: Ceng23333 <441651826@qq.com>

Signed-off-by: Ceng23333 <441651826@qq.com>

* issue/695 c++ infinicore::nn::module支持bf16
Signed-off-by: Ceng23333 <441651826@qq.com>

* embedding support f16
Signed-off-by: Ceng23333 <441651826@qq.com>

* resolve comments
Signed-off-by: Ceng23333 <441651826@qq.com>

* fix format
Signed-off-by: Ceng23333 <441651826@qq.com>

---------
Signed-off-by: Ceng23333 <441651826@qq.com>

issue/682 Parameter supports TP, fix copy_from

issue/679 缓存c++的device和python的device的对应关系，以减少互相转换的耗时

Signed-off-by: Ceng23333 <441651826@qq.com>

Signed-off-by: Ceng23333 <441651826@qq.com>

* issue/634: InfiniCore 支持InfiniLM Llama模型适配
Signed-off-by: Ceng23333 <441651826@qq.com>

* .
Signed-off-by: Ceng23333 <441651826@qq.com>

---------
Signed-off-by: Ceng23333 <441651826@qq.com>