SparseTIR: Sparse Tensor Compiler for Deep Learning
-
Updated
Mar 31, 2023 - Python
SparseTIR: Sparse Tensor Compiler for Deep Learning
Optimizing RetinaFace inference using TVM.
A small MLIR-based compiler with a custom dialect (hand-written verifier, DialectConversion lowering) and a custom loop-interchange optimization pass, lowering tensor/linalg through bufferization to LLVM for JIT/native execution.
Verified C++20 static f32 graph front end, reference interpreter, deterministic rewrites, and storage-planning core.
Tensor compiler using JIT kernel generation for ARM processors
Add a description, image, and links to the tensor-compiler topic page so that developers can more easily learn about it.
To associate your repository with the tensor-compiler topic, visit your repo's landing page and select "manage topics."