oneflow.utils.tensor.from_torch

oneflow.utils.tensor.from_torch(torch_tensor)Tensor

Create a oneflow tensor from torch tensor.

The returned tensor and torch tensor share the same memory.

Note

This function can be used in special data processing stages, torch’s some cpu ops can be used.

Parameters

input (torch.Tensor) – Input Tensor

Returns

oneflow.Tensor

For example:

import oneflow as flow
import torch

torch_t = torch.tensor([[1, 2, 3], [4, 5, 6]])
flow_t = flow.utils.tensor.from_torch(torch_t)

This feature from_torch is at Alpha Stage.