AILEARNER_L的博客

narrow可以理解为对张量的一种剪裁，且他们共用相同的storgea=torch.tensor([i for i in range(0,9)]).reshape(3,3)print(a)b=torch.narrow(input=a,dim=0,start=0,length=2)b#output:tensor([[0, 1, 2], [3, 4, 5], ...

Returns acontiguoustensor containing the same data asselftensor. Ifselftensor iscontiguous, this function returns theselftensor.普通的张良在内存中都是连续存放的，然而有些张量经过一些处理后（如permute）后就不是连续的了，此时这些张量无法用基于连...

改变张量维度排列>>> x = torch.randn(2, 3, 5)>>> x.size()torch.Size([2, 3, 5])>>> x.permute(2, 0, 1).size()torch.Size([5, 2, 3])

a=torch.randn(1,2,3)b=torch.max(a,dim=0).valuesc=torch.max(a,dim=1).valuesd=torch.max(a,dim=2).valuesb.shape=[2,3]c.shape=[1,3]d.shape=[1,2]

合并：torch.cat(inputs=(a, b), dimension=1)将a，b沿着第1维合并，dimension=-1时沿着最后一维合并。

torch.nn.functional.fold(input,output_size,kernel_size,dilation=1,padding=0,stride=1)将一组滑动局部块张量组合成一个包含这些张量的大张量。注意：Currently, only 4-D output tensors (batched image-like tensors) are support...

torch.nn.functional.unfold(input,kernel_size,dilation=1,padding=0,stride=1)从批量的输入张量中(batch_size,channel,H,W)提取滑动局部块(kernel_size的正方形）。Warning：Currently, only 4-D input tensors (batched image-li...

# 保存和加载整个模型torch.save(model_object, 'model.pkl')model = torch.load('model.pkl')# 仅保存和加载模型参数(推荐使用)torch.save(model_object.state_dict(), 'params.pkl')model_object.load_state_dict(torch.load('par...

CIFAR10transform = torchvision.transforms.Compose( [transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]) #首先torchvision输出的图片的范围都是[0,1] ...

O=(I-K+2P)/S +1

Variable:类似于一个tensor的升级版，里面包含了requires_grad,grad_fn,voliatea=Variable(torch.tensor([1]),volatile=True)b=Variable(torch.tensor([1]),requires_grad=False)voliate:当atensor的requires_grad=True后，与a相连的...