GitLab

Signed-off-by: Wuyufan <Wuyf1@sugon.com>
解决了如下问题：
1. FA 48个单测报错问题（fA输入:[B,S,H,D] reshape to [blocknums,blocksize,H,D]）.
2. 在解决FA 报错的时候一个kernel launch error.
3. norm_mlp问题暂时用rest_rng_state解决

See merge request dcutoolkit/deeplearing/TransformerEngine!77
Co-authored-by: Tangao <2205747538@qq.com>
Co-authored-by: wuyufffan <1095978552@qq.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

1,not find nvte_dgelu
2,fsdp_group is not none
3,CPUOffloadEnabled change to cpp_offload_v1
Signed-off-by: maxiao3 <maxiao3@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!74

1,Resolve out-of-bounds issues for types struct
2,Fix TestFusedCastFloat8Vectorwise test case failure
Signed-off-by: maxiao3 <maxiao3@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!73

Signed-off-by: zhaochao <zhaochao1@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!72
Co-authored-by: zhaochao <zhaochao1@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: maxiao3 <maxiao3@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!71

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

rollback activation offloading implementation

See merge request dcutoolkit/deeplearing/TransformerEngine!70
Co-authored-by: dongcl <791582849@qq.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!67

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>

Signed-off-by: wenjh <wenjh@sugon.com>