Initial support for enabling MyPy in CI

[mikepapadim]

This PR adds initial support for MyPy in the Jenkins pipeline.
At this point, it checks only python/tvm/runtime and python/tvm/auto_scheduler.

1. It adds check with MyPy in linting stage,
2. It add an option in Makefile for $ make mypy.

@jroesch @jwfromm

[tqchen]

Thanks @mikepapadim ! Can we merge the mypy check to the ci lint stage? This is mainly to simplify the pipeline. We can also push the items to https://github.com/apache/tvm/blob/main/tests/scripts/task_lint.sh

[mikepapadim]

Thanks @mikepapadim ! Can we merge the mypy check to the ci lint stage? This is mainly to simplify the pipeline. We can also push the items to https://github.com/apache/tvm/blob/main/tests/scripts/task_lint.sh

Sure, I will revert the changes and keep it in the lint stage.

[jcf94]

@mikepapadim Thanks! It's really great to have type check for python codes. While seems the CI failed on mypy but it didn't throw any error?

In my understanding, mypy requires python files to specifie data type explicitly for function parameters and return values, but we have not yet add the missing data type to these two dirs in this PR?

[mikepapadim]

@mikepapadim Thanks! It's really great to have type check for python codes. While seems the CI failed on mypy but it didn't throw any error?

In my understanding, mypy requires python files to specifie data type explicitly for function parameters and return values, but we have not yet add the missing data type to these two dirs in this PR?

@jcf94 you are right. This PR does not fix the missing data types, it is only enables MyPy.

At this point error logging is disabled ( by setting ignore_errors in mypy.ini). So, thats why at the end of the linting stage in Jenkins there is the following:

Checking MyPy Type defs in auto_scheduler package.

Success: no issues found in 17 source files

Checking MyPy Type defs in runtime package.

Success: no issues found in 13 source files

These two dirs are serving as placeholders atm.

Ideally, this PR adds the basic functionality.
Then, subsequent PRs can start incrementally to fix data types in specifics dirs and enable them.

[jcf94]

@mikepapadim Thanks, that make sense to me.

p.s. I just find a:

mypy: error: unrecognized arguments: --install-types

error in the CI log, does that need to be solved?

[mikepapadim]

@mikepapadim Thanks, that make sense to me.

p.s. I just find a:

mypy: error: unrecognized arguments: --install-types

error in the CI log, does that need to be solved?
I been looking into this.
It looks like MyPy version issue. I will try to reproduce it locally.
python/mypy#10596

[comaniac]

It's better to specify all linting/format checking tool versions (i.e., pylint, mypy, black, clang-format) to make sure the results are consistent.

[junrushao]

Finally!! Thanks for the great work!

As TensorIR is fully prepared for type annotations, we should probably check python/tvm/tir/schedule as well. CC: @Hzfengsy @tqchen

[tqchen]

@junrushao1994 we can followup with a PR

[mikepapadim]

Finally!! Thanks for the great work!

As TensorIR is fully prepared for type annotations, we should probably check python/tvm/tir/schedule as well. CC: @Hzfengsy @tqchen

@junrushao1994 #8367 Please have a look in this one