Tag: #ONNX
Kubernetes, DevOps and MLOps, written from things we ran and measured ourselves · 1 posts
Graph Compilers and Runtimes: Operator Fusion, torch.compile, ONNX Runtime, and TensorRT
You can get a 2x speedup with the exact same model, weights, and GPU just by changing how it runs. In eager execution, Python dispatches operations to the GPU one by one, writing large tensors to memory and reading them
2026-10-15 · 13 min read #비전 모델#실시간#연산 융합#torch.compile#ONNX