oneflow
oneflow copied to clipboard
Oneflow-Inc
Graph rename v2
本 pr 去掉 Block 上的 attribute 和 config
- 1、彻底避免重名问题；
- 2、去掉 block config；
import flow.nn.Graph as G
from oneflow.nn.graph import ModuleGraph
from oneflow.nn.module import Module
class AGraph(G):
def __init__(self, module: nn.Module):
super().__init__()
self.m = module
# self.m is ModuleBlock
# ModuleBlock 中有两大部分，一部分是原 module，一部分是 ModuleGraph
self.m.name // 默认取 eager 的 name
self.m.to(ModuleGraph).name // 取 ModuleGraph 的 name
self.m.to(Module) // 取得原 nn.Module
# 去掉 GraphModule 上的 config 的方法
self.m.to(ModuleGraph).set_stage(id, placement)
Fix issue: https://github.com/Oneflow-Inc/oneflow/issues/9193
另外支持 nn.Module 多重继承时的property获取
Fix issue：https://github.com/Oneflow-Inc/oneflow/issues/9345 and https://github.com/Oneflow-Inc/oneflow/issues/9186
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13965.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.3ms (= 16232.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 162.3ms / 139.7ms)
OneFlow resnet50 time: 84.9ms (= 8489.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.3ms (= 10233.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.21 (= 102.3ms / 84.9ms)
OneFlow resnet50 time: 58.1ms (= 11610.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.7ms (= 15539.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.34 (= 77.7ms / 58.1ms)
OneFlow resnet50 time: 45.2ms (= 9030.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.8ms (= 13760.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.52 (= 68.8ms / 45.2ms)
OneFlow resnet50 time: 40.3ms (= 8069.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.8ms (= 13568.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.68 (= 67.8ms / 40.3ms)
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13974.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.8ms (= 15984.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 159.8ms / 139.7ms)
OneFlow resnet50 time: 85.0ms (= 8497.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 100.5ms (= 10048.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 100.5ms / 85.0ms)
OneFlow resnet50 time: 57.7ms (= 11543.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.8ms (= 15950.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.38 (= 79.8ms / 57.7ms)
OneFlow resnet50 time: 44.5ms (= 8901.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.5ms (= 14097.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.58 (= 70.5ms / 44.5ms)
OneFlow resnet50 time: 41.7ms (= 8330.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.7ms (= 13742.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.65 (= 68.7ms / 41.7ms)
CI failed when running job: cuda-misc. PR label automerge has been removed
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13967.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.8ms (= 16077.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 160.8ms / 139.7ms)
OneFlow resnet50 time: 85.2ms (= 8519.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.3ms (= 10232.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 102.3ms / 85.2ms)
OneFlow resnet50 time: 57.6ms (= 11523.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.2ms (= 17647.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.53 (= 88.2ms / 57.6ms)
OneFlow resnet50 time: 45.3ms (= 9068.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.5ms (= 14290.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.58 (= 71.5ms / 45.3ms)
OneFlow resnet50 time: 40.1ms (= 8012.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.8ms (= 13569.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.69 (= 67.8ms / 40.1ms)
CI failed when running job: cuda-misc. PR label automerge has been removed
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.6ms (= 13956.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.8ms (= 15975.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 159.8ms / 139.6ms)
OneFlow resnet50 time: 85.3ms (= 8525.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 111.8ms (= 11184.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 111.8ms / 85.3ms)
OneFlow resnet50 time: 57.7ms (= 11532.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.7ms (= 15548.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 77.7ms / 57.7ms)
OneFlow resnet50 time: 45.2ms (= 9043.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.7ms (= 13933.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.54 (= 69.7ms / 45.2ms)
OneFlow resnet50 time: 40.5ms (= 8106.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.5ms (= 13492.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.66 (= 67.5ms / 40.5ms)
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9351/
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.6ms (= 13964.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 161.2ms (= 16115.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 161.2ms / 139.6ms)
OneFlow resnet50 time: 84.7ms (= 8471.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.9ms (= 10194.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 101.9ms / 84.7ms)
OneFlow resnet50 time: 58.0ms (= 11603.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.9ms (= 15381.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 76.9ms / 58.0ms)
OneFlow resnet50 time: 45.3ms (= 9054.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.0ms (= 13995.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.55 (= 70.0ms / 45.3ms)
OneFlow resnet50 time: 41.9ms (= 8376.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.4ms (= 13680.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.63 (= 68.4ms / 41.9ms)
CI failed when running job: cuda-misc. PR label automerge has been removed
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.9ms (= 13990.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.9ms (= 16088.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 160.9ms / 139.9ms)
OneFlow resnet50 time: 85.2ms (= 8521.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.7ms (= 10165.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.19 (= 101.7ms / 85.2ms)
OneFlow resnet50 time: 57.7ms (= 11531.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.8ms (= 15558.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 77.8ms / 57.7ms)
OneFlow resnet50 time: 45.0ms (= 9007.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.4ms (= 14273.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.58 (= 71.4ms / 45.0ms)
OneFlow resnet50 time: 39.4ms (= 7886.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.1ms (= 15425.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.96 (= 77.1ms / 39.4ms)
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13968.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.8ms (= 15983.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 159.8ms / 139.7ms)
OneFlow resnet50 time: 85.2ms (= 8524.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.1ms (= 10106.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.19 (= 101.1ms / 85.2ms)
OneFlow resnet50 time: 57.9ms (= 11581.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.9ms (= 15581.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 77.9ms / 57.9ms)
OneFlow resnet50 time: 44.6ms (= 8924.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.7ms (= 13941.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.56 (= 69.7ms / 44.6ms)
OneFlow resnet50 time: 40.6ms (= 8112.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.0ms (= 13409.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.65 (= 67.0ms / 40.6ms)
CI failed when running job: cuda-misc. PR label automerge has been removed
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13969.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.1ms (= 16409.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 164.1ms / 139.7ms)
OneFlow resnet50 time: 85.0ms (= 8495.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.0ms (= 10101.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.19 (= 101.0ms / 85.0ms)
OneFlow resnet50 time: 57.6ms (= 11515.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.2ms (= 15450.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.34 (= 77.2ms / 57.6ms)
OneFlow resnet50 time: 45.8ms (= 9153.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.4ms (= 14277.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.56 (= 71.4ms / 45.8ms)
OneFlow resnet50 time: 40.1ms (= 8026.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.9ms (= 14989.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.87 (= 74.9ms / 40.1ms)
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9351/
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.7ms (= 13972.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.7ms (= 16270.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 162.7ms / 139.7ms)
OneFlow resnet50 time: 85.1ms (= 8505.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.8ms (= 10181.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 101.8ms / 85.1ms)
OneFlow resnet50 time: 57.6ms (= 11519.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.4ms (= 15683.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.36 (= 78.4ms / 57.6ms)
OneFlow resnet50 time: 44.4ms (= 8884.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.8ms (= 14157.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.59 (= 70.8ms / 44.4ms)
OneFlow resnet50 time: 40.1ms (= 8019.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.2ms (= 14241.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.78 (= 71.2ms / 40.1ms)
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9351/
Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.
Speed stats:
GPU Name: GeForce GTX 1080
❌ OneFlow resnet50 time: 139.5ms (= 13954.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.9ms (= 15987.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 159.9ms / 139.5ms)
OneFlow resnet50 time: 85.7ms (= 8565.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.1ms (= 10314.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 103.1ms / 85.7ms)
OneFlow resnet50 time: 58.0ms (= 11592.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.7ms (= 15732.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.36 (= 78.7ms / 58.0ms)
OneFlow resnet50 time: 45.4ms (= 9072.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.7ms (= 13746.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.52 (= 68.7ms / 45.4ms)
OneFlow resnet50 time: 39.1ms (= 7829.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.0ms (= 13806.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.76 (= 69.0ms / 39.1ms)
View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9351/