oneflow.utils.tensor.to_torch

oneflow.utils.tensor.to_torch(flow_tensor)Tensor

Create a torch tensor from oneflow tensor.

The returned tensor and oneflow tensor share the same memory.

Note

Currently only local tensor is supported.

Parameters

input (oneflow.Tensor) – Input Tensor

Returns

torch.Tensor

For example:

import oneflow as flow
import torch

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

This feature to_torch is at Alpha Stage.