add randint #5718

Kevin-XiongC · 2021-08-04T03:01:47Z

python/oneflow/nn/modules/randint.py

+        >>> import numpy as np
+        >>> generator = flow.Generator()
+        >>> generator.manual_seed(0)
+        >>> flow.randint(10,(1,10),generator=generator)


oneflow/user/kernels/randint_kernel.cu

+      (user_op::HobDeviceTag() == "gpu"));
+
+REGISTER_GPU_RANDINT_KERNEL
+}  // namespace oneflow


python/oneflow/nn/modules/randint.py

+        high: flow.int32,
+        size: tuple,
+        generator: flow.Generator = None,
+        dtype: flow.dtype = flow.int32,


oneflow/user/ops/randint_op.cpp

+    .Output("out")
+    .Attr<int64_t>("low")
+    .Attr<int64_t>("high")
+    .Attr<int64_t>("seed", -1)


oneflow/user/ops/randint_op.cpp

+      return Maybe<void>::Ok();
+    });
+
+}  // namespace oneflow


python/oneflow/nn/modules/randint.py

+def randint(
+    low: flow.int64 = 0,
+    high: Union[int, tuple] = None,
+    size: tuple = None,
+    generator: flow.Generator = None,
+    dtype: flow.dtype = flow.int64,
+    layout=None,
+    device: flow.device = flow.device("cpu"),
+    requires_grad: bool = False,
+) -> flow.Tensor:


python/oneflow/nn/modules/randint.py

…into dev-randint

github-actions · 2021-08-27T05:35:44Z

CI failed, removing label automerge

…into dev-randint

github-actions · 2021-08-27T14:31:52Z

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.1ms (= 6405.2ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 141.0ms (= 7047.8ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.10 (= 141.0ms / 128.1ms)

OneFlow resnet50 time: 74.6ms (= 3730.1ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 84.5ms (= 4222.8ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.13 (= 84.5ms / 74.6ms)

OneFlow resnet50 time: 47.2ms (= 2361.6ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 58.0ms (= 2898.3ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.23 (= 58.0ms / 47.2ms)

OneFlow resnet50 time: 40.1ms (= 2007.3ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 47.8ms (= 2389.2ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.19 (= 47.8ms / 40.1ms)

OneFlow resnet50 time: 43.2ms (= 2161.7ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 44.7ms (= 2237.3ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.03 (= 44.7ms / 43.2ms)

OneFlow resnet50 time: 145.6ms (= 7278.5ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 150.5ms (= 7524.7ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.03 (= 150.5ms / 145.6ms)

OneFlow resnet50 time: 94.4ms (= 4717.7ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 93.0ms (= 4651.0ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.99 (= 93.0ms / 94.4ms)

OneFlow resnet50 time: 69.9ms (= 3493.9ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 67.1ms (= 3356.0ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.96 (= 67.1ms / 69.9ms)

OneFlow resnet50 time: 62.7ms (= 3135.5ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 56.9ms (= 2847.3ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.91 (= 56.9ms / 62.7ms)

OneFlow resnet50 time: 54.2ms (= 2710.8ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 49.5ms (= 2474.1ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.91 (= 49.5ms / 54.2ms)

Kevin-XiongC added 2 commits August 4, 2021 11:01

add randint

82859ca

Merge branch 'master' into dev-randint

c47c61e

Kevin-XiongC mentioned this pull request Aug 4, 2021

OneFlow 算子对齐 PyTorch 完备计划推进表 #4936

Closed

Kevin-XiongC marked this pull request as draft August 4, 2021 04:18

Merge branch 'master' into dev-randint

a791cde

Kevin-XiongC added api automerge feature test enhancement labels Aug 4, 2021

Kevin-XiongC added 4 commits August 4, 2021 14:22

Merge branch 'master' into dev-randint

1af2a4e

add

2d8229e

Merge branch 'master' into dev-randint

dc9c153

add doc test

ca6a0c9

Kevin-XiongC marked this pull request as ready for review August 4, 2021 09:12

Kevin-XiongC requested a review from Flowingsun007 August 4, 2021 09:25