Merge pull request !2140 from ShidaHe/debugger_dev

Merge pull request !2494 from jiangjinsheng/issue_fix4

Merge pull request !2475 from ougongchang/master

Merge pull request !2452 from ougongchang/fix_dataset_bug

Merge pull request !2468 from Giancarlo/optimizer_update

Merge pull request !2432 from liangzelang/dev1

Merge pull request !2457 from VectorSL/gpu-type-expand

Merge pull request !2180 from chenweifeng/gelu-fp16

Merge pull request !2482 from lilei/sparse_proximal_ada_grad

Merge pull request !2479 from dinghao/master

Merge pull request !2436 from jiangjinsheng/issue_fix4

Merge pull request !2434 from jjfeing/master

Merge pull request !2297 from jiangjinsheng/vm_matrixdiag

Merge pull request !2459 from wukesong/mkdify_alexnet

Merge pull request !2305 from 李鸿章/callback

Merge pull request !2466 from mindspore_ding/backend

Merge pull request !2465 from chentingting/gcn_rectification

Merge pull request !2451 from 王东旭/master

The summary operator will be optimized when it return the origin value
in constant folding scene. So I return a None value to avoid this.

Merge pull request !2474 from Chong/zc

Merge pull request !2418 from caojian05/ms_master_dev

Merge pull request !2460 from kisnwang/optimize-cpu-reduce-gradient

Merge pull request !2376 from Chong/zc

Merge pull request !2403 from JoyLvliang/session-code-review

Merge pull request !2446 from zhoufeng/xiu-ba-ge

Merge pull request !2443 from xiefangqi/xfq_fix_notice

Merge pull request !2401 from lianliguang/fix-code-review

Merge pull request !2241 from zhangzheng/accumulate

Merge pull request !2414 from limingqi107/master