Tong Chen
Tong Chen
KrnlSeqAlloc and KrnlSeqDealloc are introduced to support deep deallocation for the sequence type. The lowering of onnx sequence ops are changed according to the new design, which is described in...
This PR handles loop collapsing in op fusion. There are two parts: 1. Decide which loops to be collapsed. When ops are fused, all its operands should be checked whether...
I found that the proposal from Microsoft of using Linalg worth of exploring. The Linalg has some good features and there are existing optimization passes and backend. I am working...
Since onnx-mlir can not take string type inputs for model yet. I use the following constant string to test. ``` func.func @test_equal_string() -> tensor { %c1 = onnx.Constant dense :...
The current ONNXCallOp is used to call a function defined within ONNX model. When the function is not defined, namely is an external function, ONNXCustomOp is used. It turned out...
ONNXBroadcastOpShapeHelper is used to generate access expression for load, similar to the usage in the Element-wise op lowering. But they are different in two cases. 1. Constructor. In element-wise op...
# Bug Report ### Is the issue related to model conversion? No ### Describe the bug The shape of the output tensors for Unique Op is input content dependent. However,...
The goal is to create a python package so that we can run an inference like onnxruntime in python script. Since the RunONNXModel.py provides most of the necessary functionality, I...