PyTorch 错误 RuntimeError: view size is not compatible with input tensor‘s size and stride (at least o

最新推荐文章于 2025-01-08 18:46:09 发布

gykaula

最新推荐文章于 2025-01-08 18:46:09 发布

阅读量422

点赞数

CC 4.0 BY-SA版权

本文链接：https://blog.youkuaiyun.com/u010560584/article/details/119842584

在处理gender.py文件的第212行时遇到了一个UserWarning，关于过时的cuda使用方式。之后在尝试转换张量时遇到了RuntimeError，错误指出视图大小与输入张量的大小和步长不兼容，建议使用reshape()代替。解决方案是将'correct_k=correct[:k].view(-1)'替换为'correct_k=correct[:k].contiguous().view(-1)'。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

PyTorch 错误 RuntimeError: view size is not compatible with input tensor’s size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(…) instead.

flyfish
具体提示如下

Processinggender.py:212: UserWarning: This overload of cuda is deprecated:
	cuda(torch.device device, bool async, *, torch.memory_format memory_format)
Consider using one of the following signatures instead:
	cuda(torch.device device, bool non_blocking, *, torch.memory_format memory_format) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:882.)
  inputs, targets = inputs.cuda(), targets.cuda(async=True)
Traceback (most recent call last):

    correct_k = correct[:k].view(-1)
RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

解决方案
按照提示
correct_k = correct[:k].view(-1)
更改为
correct_k = correct[:k].contiguous().view(-1)