【pytorch】打印当前 Pytorch 保持的 tensor

最新推荐文章于 2023-03-12 15:58:18 发布

x66ccff

最新推荐文章于 2023-03-12 15:58:18 发布

阅读量258

点赞数

分类专栏： python 文章标签： pytorch python 深度学习

原文链接：https://discuss.pytorch.org/t/how-to-debug-causes-of-gpu-memory-leaks/6741

版权

python 专栏收录该内容

41 篇文章 2 订阅

订阅专栏

代码

# prints currently alive Tensors and Variables
import torch
import gc
for obj in gc.get_objects():
    try:
        if torch.is_tensor(obj) or (hasattr(obj, 'data') and torch.is_tensor(obj.data)):
            print(type(obj), obj.size())
    except:
        pass

输出

<class 'torch.Tensor'> torch.Size([9, 2])
<class 'torch.Tensor'> torch.Size([153, 2])
<class 'torch.Tensor'> torch.Size([24633, 2])
<class 'torch.Tensor'> torch.Size([25040000, 2])
<class 'torch.Tensor'> torch.Size([25040000, 2])
<class 'torch.nn.parameter.Parameter'> torch.Size([0])
<class 'torch.Tensor'> torch.Size([1, 1])
<class 'torch.Tensor'> torch.Size([1, 1])
<class 'torch.Tensor'> torch.Size([9])
<class 'torch.Tensor'> torch.Size([9, 9])
<class 'torch.Tensor'> torch.Size([9, 9])
<class 'torch.Tensor'> torch.Size([153])
<class 'torch.Tensor'> torch.Size([153, 153])
<class 'torch.Tensor'> torch.Size([153, 153])
<class 'torch.Tensor'> torch.Size([24633])
<class 'torch.Tensor'> torch.Size([5000])
<class 'torch.Tensor'> torch.Size([5000, 5000])
<class 'torch.Tensor'> torch.Size([5000, 5000])
<class 'torch.Tensor'> torch.Size([25040000])
<class 'torch.Tensor'> torch.Size([5000])
<class 'torch.Tensor'> torch.Size([5000, 5000])
<class 'torch.Tensor'> torch.Size([5000, 5000])
<class 'torch.Tensor'> torch.Size([25040000])
<class 'torch.Tensor'> torch.Size([100, 1])
<class 'torch.Tensor'> torch.Size([100])
<class 'torch.Tensor'> torch.Size([100])
<class 'torch.Tensor'> torch.Size([100, 1])
<class 'torch.Tensor'> torch.Size([100, 1])