5,323
edits
PyTorch: Difference between revisions
→Reducing memory usage
| Line 74: | Line 74: | ||
* [https://pytorch.org/docs/stable/autograd.html#torch.Tensor.detach <code>detach()</code> ]deletes the item from the autograd edge. | * [https://pytorch.org/docs/stable/autograd.html#torch.Tensor.detach <code>detach()</code> ]deletes the item from the autograd edge. | ||
* [https://pytorch.org/docs/stable/tensors.html?highlight=cpu#torch.Tensor.cpu <code>cpu()</code>] moves the tensor to the CPU. | * [https://pytorch.org/docs/stable/tensors.html?highlight=cpu#torch.Tensor.cpu <code>cpu()</code>] moves the tensor to the CPU. | ||
* [https://pytorch.org/docs/stable/tensors.html#torch.Tensor.numpy <code>numpy()</code>] returns the | * [https://pytorch.org/docs/stable/tensors.html#torch.Tensor.numpy <code>numpy()</code>] returns a numpy view of the tensor. | ||
When possible, use functions which return new views of existing tensors rather than making duplicates of tensors: | When possible, use functions which return new views of existing tensors rather than making duplicates of tensors: | ||