Code

Improving nnUNet inference speed#2048

Code

Merged

FabianIsensee merged 8 commits intoMIC-DKFZ:masterMIC-DKFZ/nnUNet:masterfrom ancestor-mithril:dev2ancestor-mithril/nnUNet:dev2Copy head branch name to clipboardApr 11, 2024

+70-73

Commits

Commits on Mar 28, 2024

• Improved data iterators

  ancestor-mithril

  authored
  d87fa5b

  Copy full SHA for d87fa5b

• Using backends.cudnn.benchmark during prediction for faster inference

  ancestor-mithril

  authored
  ea88fb7

  Copy full SHA for ea88fb7

• Using torch.inference_mode for prediction

  ancestor-mithril

  authored
  db7bec7

  Copy full SHA for db7bec7

• Removed tuple unpacking for mirroring and predicting

  ancestor-mithril

  authored
  a3e1170

  Copy full SHA for a3e1170

• Using gaussian only when necessary + applied inplace addition

  ancestor-mithril

  authored
  cf7696e

  Copy full SHA for cf7696e

• Compute gaussian

  Show description for c996ba2
  ancestor-mithril

  authored
  c996ba2

  Copy full SHA for c996ba2

• Faster compute_gaussian

  ancestor-mithril

  authored
  2d951e3

  Copy full SHA for 2d951e3

Commits on Apr 10, 2024

• Fixed after review

  Show description for 809277b
  ancestor-mithril

  committed
  809277b

  Copy full SHA for 809277b