oneflow
oneflow copied to clipboard
Fix inplace mul 0size check bug
#close https://github.com/Oneflow-Inc/oneflow/issues/9131
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9132/
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 140.0ms (= 14003.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.0ms (= 16204.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 162.0ms / 140.0ms)
OneFlow resnet50 time: 86.3ms (= 8626.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.1ms (= 10205.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 102.1ms / 86.3ms)
OneFlow resnet50 time: 58.0ms (= 11602.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.5ms (= 15698.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 78.5ms / 58.0ms)
OneFlow resnet50 time: 44.7ms (= 8944.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.4ms (= 13889.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.55 (= 69.4ms / 44.7ms)
OneFlow resnet50 time: 41.0ms (= 8208.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.9ms (= 13578.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.65 (= 67.9ms / 41.0ms)
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 140.1ms (= 14014.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.9ms (= 16087.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 160.9ms / 140.1ms)
OneFlow resnet50 time: 85.7ms (= 8572.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.9ms (= 10285.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 102.9ms / 85.7ms)
OneFlow resnet50 time: 58.5ms (= 11706.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.8ms (= 15750.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 78.8ms / 58.5ms)
OneFlow resnet50 time: 44.4ms (= 8887.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 82.2ms (= 16432.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.85 (= 82.2ms / 44.4ms)
OneFlow resnet50 time: 39.8ms (= 7959.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.0ms (= 13795.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.73 (= 69.0ms / 39.8ms)
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9132/