bump to cpp17 #9632

daquexian · 2022-12-20T04:28:36Z

升级 c++ 标准到 17，PyTorch 已经升级了 pytorch/pytorch#85969

同时废弃对 cuda 10 的支持

C++17 主要特性：

使用结构化绑定可以方便地返回多值，相比传入指针或引用的方式更清晰简洁（https://zhuanlan.zhihu.com/p/139140185 ）：
```
std::pair<int, int> f() {
  return {1, 2};
}
auto [a, b] = f();  // 等效于 a=1, b=2
```

使用 if constexpr 在很多情况下可以大大简化代码而不用再写 SFINAE（https://zhuanlan.zhihu.com/p/380856254 ）：

template<typename T>
std::string toString(T t) {
  if constexpr (std::is_same_v<T, std::string>) {
    return t;
  } else {
    return std::to_string(t);
  }
}

提供 std::***_v 代替 std::***::value
自带 std::variant 和 std::optional

本 PR 改动内容：

升级 json 库修复对 <filesystem> 头文件的检测错误
op_schema_emitter.cpp 156 行 sourceIncludeFilename 改为 sourceIncludeFilename.getValue()，修复 c++17 引入 filesystem 后 sourceIncludeFilename 类型为 cl::opt<std::string> 而引起的 ambiguous 调用
删掉 cplusplus_17.cpp/.h 文件
在 CI 中改为用 gcc9 而不是 gcc7 编译 oneflow，停止支持和 c++17 不兼容的 cuda 10.2
skip/修改了一些 test case

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-01-26T15:05:56Z

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.

Signed-off-by: daquexian <daquexian566@gmail.com>

.github/workflows/test.yml

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-01-27T08:04:29Z

Speed stats:

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-01-27T11:53:44Z

Speed stats:

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-01-27T12:18:06Z

Speed stats:

This reverts commit f8a416d.

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-02-04T12:12:00Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

This reverts commit afea0f6.

….py to fix oom Signed-off-by: daquexian <daquexian566@gmail.com>

…vative" This reverts commit 5bc7fa0.

github-actions · 2023-02-04T13:40:59Z

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 153.0ms (= 15302.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.6ms (= 16459.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 164.6ms / 153.0ms)

OneFlow resnet50 time: 92.9ms (= 9287.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.2ms (= 10318.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.2ms / 92.9ms)

OneFlow resnet50 time: 60.8ms (= 12159.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.2ms (= 15832.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 79.2ms / 60.8ms)

OneFlow resnet50 time: 42.6ms (= 8514.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.5ms (= 15894.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.87 (= 79.5ms / 42.6ms)

OneFlow resnet50 time: 36.3ms (= 7263.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.1ms (= 13822.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.90 (= 69.1ms / 36.3ms)

github-actions · 2023-02-04T13:51:36Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

github-actions · 2023-02-04T18:27:11Z

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 153.0ms (= 15296.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.3ms (= 16434.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.3ms / 153.0ms)

OneFlow resnet50 time: 92.8ms (= 9279.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.2ms (= 10423.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 104.2ms / 92.8ms)

OneFlow resnet50 time: 60.4ms (= 12079.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 86.9ms (= 17387.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.44 (= 86.9ms / 60.4ms)

OneFlow resnet50 time: 42.4ms (= 8480.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.4ms (= 14476.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.71 (= 72.4ms / 42.4ms)

OneFlow resnet50 time: 36.4ms (= 7281.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.2ms (= 14030.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.93 (= 70.2ms / 36.4ms)

github-actions · 2023-02-05T02:08:38Z

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 152.8ms (= 15282.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.0ms (= 16300.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 163.0ms / 152.8ms)

OneFlow resnet50 time: 91.8ms (= 9178.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 105.6ms (= 10557.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 105.6ms / 91.8ms)

OneFlow resnet50 time: 60.5ms (= 12106.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 82.4ms (= 16485.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.36 (= 82.4ms / 60.5ms)

OneFlow resnet50 time: 42.9ms (= 8584.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.5ms (= 14301.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.67 (= 71.5ms / 42.9ms)

OneFlow resnet50 time: 37.3ms (= 7459.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 82.1ms (= 16426.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 2.20 (= 82.1ms / 37.3ms)

github-actions · 2023-02-05T02:12:14Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-02-05T05:03:55Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

github-actions · 2023-02-05T05:04:38Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14130.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.6ms (= 14457.7ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.02 (= 144.6ms / 141.3ms)

OneFlow resnet50 time: 82.0ms (= 8199.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.0ms (= 8698.9ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.06 (= 87.0ms / 82.0ms)

OneFlow resnet50 time: 51.3ms (= 10264.4ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.3ms (= 12060.5ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.17 (= 60.3ms / 51.3ms)

OneFlow resnet50 time: 33.5ms (= 6691.4ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.4ms (= 8681.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.30 (= 43.4ms / 33.5ms)

OneFlow resnet50 time: 25.7ms (= 5134.3ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 44.4ms (= 8882.2ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.73 (= 44.4ms / 25.7ms)

OneFlow swin dataloader time: 0.237s (= 47.463s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 29.903s / 200, num_workers=1)
Relative speed: 0.630 (= 0.150s / 0.237s)

OneFlow swin dataloader time: 0.064s (= 12.756s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.439s / 200, num_workers=4)
Relative speed: 0.662 (= 0.042s / 0.064s)

OneFlow swin dataloader time: 0.037s (= 7.331s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.601s / 200, num_workers=8)
Relative speed: 0.628 (= 0.023s / 0.037s)

❌ OneFlow resnet50 time: 153.0ms (= 15299.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.3ms (= 16433.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.3ms / 153.0ms)

OneFlow resnet50 time: 93.1ms (= 9310.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.7ms (= 10373.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.7ms / 93.1ms)

OneFlow resnet50 time: 61.4ms (= 12281.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.8ms (= 15952.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 79.8ms / 61.4ms)

OneFlow resnet50 time: 42.6ms (= 8513.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.1ms (= 14212.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.67 (= 71.1ms / 42.6ms)

OneFlow resnet50 time: 38.3ms (= 7659.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.1ms (= 13822.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.80 (= 69.1ms / 38.3ms)

github-actions · 2023-02-05T05:47:14Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.2ms (= 14116.1ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.8ms (= 14184.7ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.00 (= 141.8ms / 141.2ms)

OneFlow resnet50 time: 81.6ms (= 8157.8ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.0ms (= 8696.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 87.0ms / 81.6ms)

OneFlow resnet50 time: 50.8ms (= 10157.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.5ms (= 12103.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.19 (= 60.5ms / 50.8ms)

OneFlow resnet50 time: 33.9ms (= 6778.6ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 46.6ms (= 9328.5ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.38 (= 46.6ms / 33.9ms)

OneFlow resnet50 time: 25.4ms (= 5079.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 41.6ms (= 8325.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.64 (= 41.6ms / 25.4ms)

OneFlow swin dataloader time: 0.233s (= 46.602s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 30.017s / 200, num_workers=1)
Relative speed: 0.644 (= 0.150s / 0.233s)

OneFlow swin dataloader time: 0.065s (= 12.973s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.251s / 200, num_workers=4)
Relative speed: 0.636 (= 0.041s / 0.065s)

OneFlow swin dataloader time: 0.036s (= 7.292s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.528s / 200, num_workers=8)
Relative speed: 0.621 (= 0.023s / 0.036s)

❌ OneFlow resnet50 time: 153.1ms (= 15307.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.5ms (= 16352.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 163.5ms / 153.1ms)

OneFlow resnet50 time: 93.9ms (= 9386.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 110.9ms (= 11086.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 110.9ms / 93.9ms)

OneFlow resnet50 time: 61.4ms (= 12273.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 85.4ms (= 17074.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.39 (= 85.4ms / 61.4ms)

OneFlow resnet50 time: 42.2ms (= 8434.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.2ms (= 14230.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.69 (= 71.2ms / 42.2ms)

OneFlow resnet50 time: 35.9ms (= 7178.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.7ms (= 15748.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 2.19 (= 78.7ms / 35.9ms)

github-actions · 2023-02-05T05:47:21Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

github-actions · 2023-02-05T06:15:51Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14129.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.0ms (= 14401.9ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.02 (= 144.0ms / 141.3ms)

OneFlow resnet50 time: 82.5ms (= 8248.7ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.7ms (= 8772.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.06 (= 87.7ms / 82.5ms)

OneFlow resnet50 time: 51.0ms (= 10200.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.0ms (= 12008.1ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.18 (= 60.0ms / 51.0ms)

OneFlow resnet50 time: 34.1ms (= 6828.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.5ms (= 8693.1ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.27 (= 43.5ms / 34.1ms)

OneFlow resnet50 time: 27.1ms (= 5412.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 35.2ms (= 7042.1ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.30 (= 35.2ms / 27.1ms)

OneFlow swin dataloader time: 0.237s (= 47.385s / 200, num_workers=1)
PyTorch swin dataloader time: 0.153s (= 30.589s / 200, num_workers=1)
Relative speed: 0.646 (= 0.153s / 0.237s)

OneFlow swin dataloader time: 0.064s (= 12.776s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.491s / 200, num_workers=4)
Relative speed: 0.665 (= 0.042s / 0.064s)

OneFlow swin dataloader time: 0.037s (= 7.437s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.439s / 200, num_workers=8)
Relative speed: 0.597 (= 0.022s / 0.037s)

❌ OneFlow resnet50 time: 153.3ms (= 15328.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.2ms (= 16422.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.2ms / 153.3ms)

OneFlow resnet50 time: 92.8ms (= 9278.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 112.6ms (= 11256.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.21 (= 112.6ms / 92.8ms)

OneFlow resnet50 time: 61.1ms (= 12228.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.3ms (= 15867.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 79.3ms / 61.1ms)

OneFlow resnet50 time: 43.4ms (= 8672.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 82.1ms (= 16424.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.89 (= 82.1ms / 43.4ms)

OneFlow resnet50 time: 35.3ms (= 7065.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.9ms (= 13772.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.95 (= 68.9ms / 35.3ms)

github-actions · 2023-02-05T06:18:29Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

Signed-off-by: daquexian <daquexian566@gmail.com>

github-actions · 2023-02-05T07:48:24Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.2ms (= 14124.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 140.9ms (= 14093.0ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.00 (= 140.9ms / 141.2ms)

OneFlow resnet50 time: 82.6ms (= 8255.6ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 89.8ms (= 8980.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.09 (= 89.8ms / 82.6ms)

OneFlow resnet50 time: 51.1ms (= 10215.6ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.5ms (= 11892.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.16 (= 59.5ms / 51.1ms)

OneFlow resnet50 time: 34.0ms (= 6801.6ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 46.3ms (= 9263.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.36 (= 46.3ms / 34.0ms)

OneFlow resnet50 time: 25.7ms (= 5141.5ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 43.6ms (= 8720.0ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.70 (= 43.6ms / 25.7ms)

OneFlow swin dataloader time: 0.237s (= 47.441s / 200, num_workers=1)
PyTorch swin dataloader time: 0.153s (= 30.622s / 200, num_workers=1)
Relative speed: 0.645 (= 0.153s / 0.237s)

OneFlow swin dataloader time: 0.064s (= 12.804s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.249s / 200, num_workers=4)
Relative speed: 0.644 (= 0.041s / 0.064s)

OneFlow swin dataloader time: 0.037s (= 7.499s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.567s / 200, num_workers=8)
Relative speed: 0.609 (= 0.023s / 0.037s)

❌ OneFlow resnet50 time: 153.2ms (= 15316.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.2ms (= 16421.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.2ms / 153.2ms)

OneFlow resnet50 time: 93.2ms (= 9321.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 106.3ms (= 10627.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 106.3ms / 93.2ms)

OneFlow resnet50 time: 61.4ms (= 12283.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.0ms (= 15790.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 79.0ms / 61.4ms)

OneFlow resnet50 time: 42.6ms (= 8511.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.2ms (= 14244.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.67 (= 71.2ms / 42.6ms)

OneFlow resnet50 time: 36.4ms (= 7280.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 85.7ms (= 17148.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 2.36 (= 85.7ms / 36.4ms)

github-actions · 2023-02-05T07:52:14Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

github-actions · 2023-02-05T08:52:16Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9632/

[升级 c++17](#9632) 之后，把 `ParseDeviceString` 和 `ParseDeviceNameConf` 从修改指针改为返回多值 Signed-off-by: daquexian <daquexian566@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

bump to cpp17

41b740b

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian marked this pull request as ready for review December 20, 2022 04:28

daquexian requested a review from jackalcooper as a code owner December 20, 2022 04:28

daquexian changed the title ~~bump to cpp17~~ [don't merge] bump to cpp17 Dec 20, 2022

daquexian added feature build system labels Dec 20, 2022

daquexian requested a review from oneflow-ci-bot December 20, 2022 04:29

daquexian added 4 commits December 20, 2022 12:37

use cu112 instead of cu102 in ci

da46198

Signed-off-by: daquexian <daquexian566@gmail.com>

add missing header

c4d49b4

Signed-off-by: daquexian <daquexian566@gmail.com>

Merge branch 'master' into cpp17

9f73684

Signed-off-by: daquexian <daquexian566@gmail.com>

upgrade nlohmann/json for better c++17 support

c59ccc1

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian requested review from oneflow-ci-bot and removed request for oneflow-ci-bot January 4, 2023 09:34

daquexian and others added 2 commits January 26, 2023 23:04

use gcc9 instead of gcc7 for better c++17 support

7962f29

Signed-off-by: daquexian <daquexian566@gmail.com>

auto format by CI

759836f

daquexian requested review from oneflow-ci-bot and removed request for oneflow-ci-bot January 26, 2023 16:03

Merge branch 'master' into cpp17

8c6a86c

Signed-off-by: daquexian <daquexian566@gmail.com>

jackalcooper approved these changes Jan 27, 2023

View reviewed changes

.github/workflows/test.yml Outdated Show resolved Hide resolved

daquexian added 2 commits January 27, 2023 14:53

fix ambiguous function call in c++17

fe413ab

Signed-off-by: daquexian <daquexian566@gmail.com>

CI: cu112 -> cu117

7cc8d26

Signed-off-by: daquexian <daquexian566@gmail.com>

update test.yml for correct compute-platform

f8a416d

Signed-off-by: daquexian <daquexian566@gmail.com>

use get-oneflow torch1.13 branch

6750bbf

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian added 2 commits January 27, 2023 20:25

Revert "update test.yml for correct compute-platform"

542bd7d

This reverts commit f8a416d.

use cuda 11.6 temporarily

75b6183

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian added 3 commits February 4, 2023 20:53

Revert "decrease parallel num of 1n4d tests from 4 to 3"

c7bfe29

This reverts commit afea0f6.

add missing skip_unless_1n1d and use unittest.main() in test_swautils…

6771c2d

….py to fix oom Signed-off-by: daquexian <daquexian566@gmail.com>

Revert "manually run gc after each test in TestGlobalMathOpHigherDeri…

faaf30f

…vative" This reverts commit 5bc7fa0.

daquexian mentioned this pull request Feb 4, 2023

add swa_utils modules #9781

Merged

Merge branch 'master' into cpp17

f8200c3

daquexian added 2 commits February 5, 2023 11:06

Merge branch 'master' into cpp17

ca6587d

update minimum gcc version from 7 to 9

3a34187

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian enabled auto-merge (squash) February 5, 2023 05:29

jackalcooper added the need-all-tests-even-fail label Feb 5, 2023

reduce memory usage of TestRNNUtils

83f06f9

Signed-off-by: daquexian <daquexian566@gmail.com>

daquexian added the automerge label Feb 5, 2023

daquexian merged commit 9b5a0fb into master Feb 5, 2023

daquexian deleted the cpp17 branch February 5, 2023 09:09

daquexian mentioned this pull request Feb 6, 2023

make some functions return multi values #9833

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

bump to cpp17 #9632

bump to cpp17 #9632

daquexian commented Dec 20, 2022 •

edited

Loading

github-actions bot commented Jan 26, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

bump to cpp17 #9632

bump to cpp17 #9632

Conversation

daquexian commented Dec 20, 2022 • edited Loading

github-actions bot commented Jan 26, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Jan 27, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 4, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

github-actions bot commented Feb 5, 2023

daquexian commented Dec 20, 2022 •

edited

Loading