How does torch.compile() achieve massive speedups despite highly optimized torch functions? See how operator fusion works!
-
Updated
Aug 3, 2026 - Jupyter Notebook
#
operator-fusion
Here are 11 public repositories matching this topic...
面向 Transformer 推理的可验证 Triton 融合算子与 torch.library 集成
python acceleration deep-learning high-performance cuda inference pytorch nvidia stable triton gpu-computing gpu-kernels kernel-fusion ai-infra llm rotary-position-embedding rmsnorm llm-inference operator-fusion transformer-inference
-
Updated
Oct 19, 2026 - Python
A specialized compiler that optimizes deep learning models for AI accelerators with operator fusion, memory optimization, and hardware-specific passes.
-
Updated
Mar 1, 2026 - Python
C++17 ONNX inference optimizer + CPU runtime for Apple Silicon. Operator fusion via IR passes, Accelerate AMX-backed sgemm; benched vs ONNX Runtime CPU EP on DistilBERT (1.26x baseline speedup, 6.99x ORT on raw MatMul).
machine-learning protobuf compiler cpp transformer accelerate cpp17 amx inference-engine sgemm ml-infrastructure onnx gelu distilbert apple-silicon operator-fusion
-
Updated
Sep 24, 2026 - C++
Profiling-driven micro-architectural case study & custom Triton GPU operators for batch-1 edge vision inference on NVIDIA Ada Lovelace.
performance-engineering computer-vision cuda pytorch yolo triton gpu-acceleration microarchitecture ada-lovelace edge-ai operator-fusion nsight-compute kernel-optimization amdahls-law
-
Updated
Aug 28, 2026 - Python
TensorMorph is an AI-assisted MLIR compiler for TOSA graph optimization and operator fusion.
compiler graph-optimization mlir operator-fusion ml-compiler tosa tensor-optimization ai-assisted-compilation
-
Updated
Jan 12, 2026 - C++
Native Rust edge inference engine with zero-copy memmap2 tensor loading, register-fused Linear+ReLU kernels, and scenario-aware MoE routing via rayon work-stealing — achieving 352µs lightweight and 1.39ms dense expert execution.
rust high-performance zero-copy parallel-computing systems-programming inference-engine edge-computing memory-mapping mixture-of-experts operator-fusion
-
Updated
May 31, 2026 - Rust
Small CPU graph compiler: constant folding, dead-code removal, fused vs unfused C lowering, opt-in CSE. Benchmarked, with limits stated. AI-assisted.
-
Updated
Oct 17, 2026 - Python
From-scratch C++ inference compiler: ONNX parsing, operator fusion, and handwritten NCHW kernels that match ONNX Runtime exactly on a trained MNIST CNN.
-
Updated
Jul 8, 2026 - C++
A small ML graph compiler: ONNX IR, fusion and folding passes (141 nodes to 32 on resnet18, outputs identical to 1.2e-06), liveness memory planning (13.8MB to 4MB), NumPy runtime verified against onnxruntime
-
Updated
Aug 29, 2026 - Python
A deep learning compiler with operator fusion and GPU codegen. PyTorch to graph IR to Triton and CUDA C++, with an analytic cost model checked against measured A100 speedups.
-
Updated
Aug 28, 2026 - Python
Add this topic to your repo
To associate your repository with the operator-fusion topic, visit your repo's landing page and select "manage topics."