Tong Chen

ConstantOp may have attribute other than value_sparse or value_value, for instant value_int, value_ints and etc. I remember I wrote a bunch of transformation rules to normalize those attribute to value_value...

Thanks. Pass ordering problem. Will we have the problem in the new hybrid transformation? I will try to add a canonicalization pass before shape inference.

> I tried adding canonicalizing before shape inference with the patch in the previous message, on top of the fixes in PR #2232 and now `onnx-mlir gpt2-dataprop.onnx` fails with these...

@sorenlassen By the way, this model has lots of custom Ops. We can use it as a test case.

Adding a canonicalization pass caused the numerical test to fail. The error message is ``` error: type of return operand 0 ('tensor

@sorenlassen I went back to review this PR. I agree on the basic idea to define the hybrid pass to replace most of the onnx-to-onnx transformations. Several questions 1. When...

@sorenlassen I thought over the problem further. The main obstacle is from LoopOp. The shape inference for loop body starts with assigning the type of loop initial values to the...

Comment on code: The major change is to decompose the current inferShape function for control flow ops, IfOp and LoopOp,into two pieces. One passes the shape info into the region...

> @chentong319 if you want to discuss in person I can meet Friday morning March 17 anytime between 8-11am PT if you're available then Sorry, I am not available this...

> > Does any time next Monday work for you? > > yes, I'm available anytime after 1 PT on Monday 1 pt Monday https://ibm.webex.com/meet/chentong90