* Remove 'final' in GraphExecutor for extension

* Dynamic num of inputs/outputs for tvm_op

* [REFACTOR/PASS] Formalize argument bind and match util

* grammar

* [DOCS] fix doc builder

* fix

* fix

* fix doc builder

* Add NNPack Fully Connected Inference

* Add NNPack fully_connected_output

* Fix lint

* Fix

* [INTRINSIC] Add sqrt

* [INTRINSIC] Expose on cpp

* [OP] Initial start of op library

* add gtest

* [LANG] Add all and any in the python API

* compatible with python3

* [LANG] Expose tvm.cast

* Update

* Add unittest

* [CODEGEN] Make codegen registerable

* fix llvm disbaled

* [CODEGEN] More storage alignment info aware generation

* fix

* fix

* fix warning

* [RUNTIME] Move device_api to include

* fix doxygen

* fix device api

* fx

* [TUTORIAL] Add tutorial for RPC

* [TUTORIAL] Update tutorial

* [TUTORIAL] Update tutorial

* trigger update

* [TUTORIAL] Improve build

* Support for batch ComputeOp

* Support for batch ComputeOp

* Fix CrossThreadReduction

* Fix lint

* Add UpdateArray, remove support for batch reduce

* Tuple input support for reduce

* rfactor works with multiple reducer; support multiple reducers with different types

* Small fix

* Small fix

* Change return type of rfactor to Array<Expr>

* Fix lint

* Improve

* Add tutorial

* Improve tutorial

* Improve tutorial

* Change Schedule Array constructor to static make method

* Add CreateSchedule

* Add doc

* Change CreateSchedule to create_schedule at cpp side

* [BUILD/CODEGEN] Allow combine multiple functions in build stage.

* Enhance code module

* fix compile

* [PERF/TIMER] Add buildin timing logic

* fix lint

* Improve makefile

* Fix