[examples][MLIRPython] support different tensor dtype in importer #180
Conversation
xTayEx
force-pushed
the
feat/different_tensor_dtype_importer
branch
from
e98ee65
to
e3b438b
Compare
xTayEx
force-pushed
the
feat/different_tensor_dtype_importer
branch
from
e3b438b
to
39ca419
Compare
Got it! I have add some test and use pattern match syntax in the new update. Give a review when you're free. Thanks :) |
There was a problem hiding this comment.
It looks good. There are a few other things that need to be done here, you can look at matmul here, the function that generates matmul should also support the type you added (i32), and examples should be provided accordingly.test_different_type.py is fuzzy,arith_add.py describes the contents stored in this file.Finally, thank you for your work .
|
When an issue is resolved, you can click |
…corresponding test
|
|
||
| int32_in1 = torch.randint(0, 10, (10,)).to(torch.int32) | ||
| int32_in2 = torch.randint(0, 10, (10,)).to(torch.int32) | ||
| foo_mlir(int32_in1, int32_in2) |
There was a problem hiding this comment.
You should not create a new test directory here, you should write the code here to arith_add.py.
There was a problem hiding this comment.
Thank you for you review. I think the arith_add.py file is for demo purpose, not for testing. Therefore I create the test_different_dtype.py for just for testing the different dtype support feature. What do you think?
There was a problem hiding this comment.
I see what you mean, but I don't think it makes sense to recreate a directory for testing purposes, and at the same time it would cause confusion in the directory structure. arith_add.py should be used as an example and also for testing purposes.I think this is an obvious thing to do.
There was a problem hiding this comment.
Got it. I have removed the test directory and put test code in arith_add.py in the latest commit. The redundant end lines should also have been removed. Give the updates a review when you're free. Thanks! :)
| def foo(x, y): | ||
| return x + y | ||
|
|
||
|
|
There was a problem hiding this comment.
The two new lines added in rows 5 and 9 look unnecessary.
There was a problem hiding this comment.
Thank you for pointing that out. I'd like to clarify that the two new lines added in rows 5 and 9 were automatically inserted by the yapf formatter, which use the .style.yapf file as configuration. If this is an issue, please let me know how you'd like me to proceed, or if there are other formatting rules I should follow.
There was a problem hiding this comment.
OK, it doesn't look like it needs to be changed.
|
@zhanghb97 I have reviewed the code as well, please review the code. |
|
Thanks! @xTayEx @linuxlonelyeagle |
…ddy-compiler#180) * [examples][MLIRPython] support different tensor dtype in importer * [examples][MLIRPython] use pattern match syntax for dtype matching
Currently, importer only support tensor with
torch.float32dtype. This PR adds support for different tensor dtype.