torch.compile benchmarks.csv

,,A100 (batch size 1),,A100 (batch size 4),,A100 (batch size 16),
task,model,with torch.compile,without torch.compile,with torch.compile,without torch.compile,with torch.compile,without torch.compile
image classification,ViT,7.584,9.325,14.499,14.832,40.01,40.944
image classification,BeiT,8.448,11.282,13.048,13.205,41.048,41.854
image classification,ConvNeXT,10.208,10.41,21.631,22.94,75.561,82.258
image classification,ResNet,4.124,6.531,4.268,6.657,5.024,7.018
object detection,DETR,19.04,34.619,32.418,48.657,161.902,164.382
object detection,OwlViT,18.42,24.978,,,,
object detection,Resnet-101,18.197,36.276,44.998,68.101,155.021,232.887
object detection,Conditional-DETR,17.993,31.219,35.552,56.47,124.032,180.491
image segmentation,Segformer,10.5,11.759,16.476,18.838,31.144,37.005
image segmentation,Mask2former,49.117,60.188,61.781,74.277,154.814,178.945
image segmentation,Maskformer,59.487,75.764,159.116,180.7,579.826,638.57
image segmentation,MobileNet,3.974,8.583,8.515,14.174,30.31,51.693