Merge pull request !4547 from ling/deconv

Merge pull request !4574 from xutianchun/quant_st

Merge pull request !4353 from lyvette/tflite_parser

Merge pull request !4556 from 张学同/to_merge

Merge pull request !4568 from cjh9368/fp16

Merge pull request !4377 from ZPaC/master-multi-server-weight-slice

Merge pull request !4510 from VectorSL/fix-codex

Merge pull request !4198 from thlinh/dev_Aug08_eliminate_more_redundant_Max

append ut

refactor tflite parsers

modify tflite parser, ut and model

supplement caffe flatten parser

fix the weight tensor format of deconv bug

fix bug when idx=-1

fix the weight tensor format of depthConv bug.

Merge pull request !4423 from lianliguang/add-reshape-type-to-tensor

Merge pull request !4554 from wangchangkai/master

Merge pull request !4565 from xutianchun/quant_0817

Merge pull request !4466 from looop5/akg_readme

1. Use C++ native bool type for tensor:
Since we changed tensor data from vector<T> to unique_ptr<T[]>,
boolean tensor can be implemented using C++ native bool type.

2. Improve ToString():
Try to make tensor string representation looks better.

Merge pull request !4560 from chenjianping/lite_dev2

Merge pull request !4482 from yao_yf/add_field_in_strategy_ckpt

Merge pull request !4512 from YuJianfeng/internal_output

Merge pull request !4527 from zhaizhiqiang/master

Merge pull request !4553 from wandongdong/up

Merge pull request !2948 from amongo/FixControlFlow

Merge pull request !4552 from lixian/master

Merge pull request !4551 from lixian/master

Merge pull request !4545 from yangruoqi713/lite

Merge pull request !4504 from wangshaocong/lite