Search results

ONNX Runtime GPU is ~2x slower than PyTorch for cellpose model inference

Status: Open.

#3 In ryouchinsa/cellpose-cpp;

· Endeavor-0808 opened on Mar 16, 2026

"On Windows 10, with device set to cuda:0, no valid mask is produced; however, normal results can be obtained when running on the CPU."

Status: Open.

#1 In ryouchinsa/cellpose-cpp;

· sunanlin13174 opened on Jul 16, 2025