Merge pull request !3550 from JesseKLee/merge_op_fix

Merge pull request !3449 from MahdiRahmaniHanzaki/cifar100-c-api

Merge pull request !3508 from ms_yan/vgg_gpu

merge the script

optimize the script

Merge pull request !3529 from ZPaC/pass-optimizer-attrs-to-push

Merge pull request !3519 from anzhengqi/remove-log

Merge pull request !3538 from LiHongzhang/f50_t20

Merge pull request !3534 from chenfei_mindspore/release-make-tuple-of-load-input-data

Merge pull request !3526 from dinghao/master

Merge pull request !3484 from caifubi/data-dump

Merge pull request !3516 from hexia/fix_python_import

Merge pull request !3536 from fangzehua/argmax

Merge pull request !3500 from lirongzhen1/master

Merge pull request !3481 from LiHongzhang/limit_summary

Merge pull request !3468 from ZhangQinghua/master1

Merge pull request !3525 from hewei/fix_parameter_bug

When `collect_tensor_freq` is specified as `None`,
the `collect_tensor_freq` would be auto calculated.

The previous behavior is to collect at most 50 steps,
now changing to 20

Merge pull request !3511 from linqingke/new_ops

1. Parameter's init_data() should have no effect if default_input already set;
2. This bug is introduced by 'decouple ParamValue from python';
3. An unit test case is added to ensure the right init_data() behavior.

Merge pull request !3513 from ZPaC/master-add-ps-test-cases

Merge pull request !3425 from fangzehua/avgpool

Merge pull request !3444 from zhouyuanshen/master

Merge pull request !3497 from yuchaojie/transformer_st

Merge pull request !3349 from liubuyu/master

Merge pull request !3490 from yangyongjie/master

Merge pull request !3495 from chenweifeng/strided_slice

Merge pull request !3324 from fangzehua/reduce_any