kimm240

How to implement sliding window with persistent line buffers using input_buffer in AI Engine?

1

Hello, I’m working on a stencil-based AI Engine kernel where I want to implement a sliding window using local memory. My goal is to reuse previously loaded rows and update...

[TIR][Schedule] FuseReductionEpilogue: Add Clipping pattern support

4

Currently, the FuseReductionEpilogue primitive only supports Bias (addition) and BiasReLU (addition + ReLU) epilogue patterns. However, clipping operations (min(max(x, lower), upper)) are commonly used in deep learning models and would...

[TIR][Schedule] FuseReductionEpilogue: Add ReLU support

2

The FuseReductionEpilogue primitive currently supports fusing bias addition epilogues into reduction blocks. This commit extends the primitive to also support ReLU activation functions in epilogue blocks, enabling fusion of patterns...

[Optimization][Operator] Implement and enable Conv2d-Reshape-Add-ReLU fusion

2

This commit extends the make_fused_bias_activation_pattern function to support PyTorch frontend's specific IR generation pattern for convolution operations with bias. When PyTorch models with bias=True are converted to Relax IR, the...

This PR introduces an operator fusion for the common `conv2d` followed by `reshape`, `add`, and `relu` sequence, commonly found in deep learning models (e.g., convolution + bias + activation pattern...

ONNX To LLVM Pass Backend using linalg

5

## Overview The Linalg pipeline transforms ONNX models into executable LLVM IR through a series of dialect conversions and optimizations. This document provides a detailed breakdown of all passes applied...