Tong Chen

@codemzs I saw that BF16 is already allowed for some ops in our current onnx dialect definition. BF16 are added for some ops, such as LeakyRelu, Scan, LessOrEqual and etc....

I am moving to v1.11.0. The changes in operations w.r.t. to the version supported previously by onnx-mlir are: ``` Add has a newer version 14 over old version 13 BatchNormalization...

@philass I went over the changes in your PR #1373. Glad that you are working on new ops introduced in the recent releases. onnx-mlir uses onnx package not only for...

No change needed for lowering. The only change of BatchNormalization is that ReduceMean has to be float when the input is float16. That is used only in the train mode....

Thanks @amanwithfruit . I can not give you that model directly. I asked the owner to simplify the model so that the buggy part can be shared.

By the way, the code handling axes for SliceOp may be used by other operations, such as ReduceSum, Squeeze and Unsqueezed, all of which changes their axes from attribute to...

I am thinking of adding a function to generate the default value in the Op definition, hook it with the optional input with default value, and the constructor invokes that...

@sorenlassen I'd like to provide you info about how optional arguments for onnx operations are supported in onnx-mlir, if the Optional Type is related. In older version of onnx, arguments...

Changes for the control 1. If the termination condition is NOT modified in loop, we can still use affine loop, but move the scf.if out. ``` %11 = krnl.load **%5**[]...

> I tried to get location information for resnet-50 model in onnx-model-zoo. > > **Command:** onnx-mlir --preserveBitcode --preserveLocations --preserveMLIR /external/onnx-zoo/models/vision/classification/resnet/model/resnet50-caffe2-v1-7.onnx > > **Expect:** > > 1. User can get mlir...