{"id":2471,"date":"2025-05-12T17:08:36","date_gmt":"2025-05-12T09:08:36","guid":{"rendered":"https:\/\/thereisno.top\/?p=2471"},"modified":"2025-05-12T17:08:36","modified_gmt":"2025-05-12T09:08:36","slug":"torch-%e5%bc%a0%e9%87%8f","status":"publish","type":"post","link":"https:\/\/thereisno.top\/?p=2471","title":{"rendered":"torch \u5f20\u91cf"},"content":{"rendered":"\n

\u5f20\u91cf\u662f\u4e00\u79cd\u7279\u6b8a\u7684\u6570\u636e\u7ed3\u6784\uff0c\u4e0e\u6570\u7ec4\u548c\u77e9\u9635\u975e\u5e38\u76f8\u4f3c\u3002\u5728 PyTorch \u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u5f20\u91cf\u6765\u7f16\u7801\u6a21\u578b\u7684\u8f93\u5165\u548c\u8f93\u51fa\uff0c\u4ee5\u53ca\u6a21\u578b\u7684\u53c2\u6570\u3002<\/p>\n\n\n\n

\u5f20\u91cf\u7c7b\u4f3c\u4e8e NumPy \u7684<\/a> ndarray\uff0c\u4e0d\u540c\u4e4b\u5904\u5728\u4e8e\u5f20\u91cf\u53ef\u4ee5\u5728 GPU \u6216\u5176\u4ed6\u786c\u4ef6\u52a0\u901f\u5668\u4e0a\u8fd0\u884c\u3002\u5b9e\u9645\u4e0a\uff0c\u5f20\u91cf\u548c NumPy \u6570\u7ec4\u901a\u5e38\u53ef\u4ee5\u5171\u4eab\u5e95\u5c42\u5185\u5b58\uff0c\u4ece\u800c\u65e0\u9700\u590d\u5236\u6570\u636e\uff08\u8be6\u89c1\u4e0e NumPy \u7684\u6865\u63a5<\/a>\uff09\u3002\u5f20\u91cf\u8fd8\u9488\u5bf9\u81ea\u52a8\u5fae\u5206\u8fdb\u884c\u4e86\u4f18\u5316\uff08\u6211\u4eec\u5c06\u5728\u540e\u9762\u7684 Autograd<\/a> \u90e8\u5206\u8be6\u7ec6\u4ecb\u7ecd\uff09\u3002\u5982\u679c\u60a8\u719f\u6089 ndarrays\uff0c\u60a8\u4f1a\u5f88\u5feb\u9002\u5e94 Tensor API\u3002\u5982\u679c\u4e0d\u719f\u6089\uff0c\u8bf7\u7ee7\u7eed\u9605\u8bfb\uff01<\/p>\n\n\n\n

import <\/strong>torch<\/strong>
import <\/strong>numpy <\/strong>as <\/strong>np<\/strong><\/p>\n\n\n\n

\u521d\u59cb\u5316\u5f20\u91cf<\/p>\n\n\n\n

\u5f20\u91cf\u53ef\u4ee5\u901a\u8fc7\u591a\u79cd\u65b9\u5f0f\u8fdb\u884c\u521d\u59cb\u5316\u3002\u8bf7\u770b\u4ee5\u4e0b\u793a\u4f8b<\/p>\n\n\n\n

\u76f4\u63a5\u4ece\u6570\u636e\u521b\u5efa<\/strong><\/p>\n\n\n\n

\u5f20\u91cf\u53ef\u4ee5\u76f4\u63a5\u4ece\u6570\u636e\u521b\u5efa\u3002\u6570\u636e\u7c7b\u578b\u4f1a\u81ea\u52a8\u63a8\u65ad\u3002<\/p>\n\n\n\n

data = [[<\/strong>1, <\/strong>2],[<\/strong>3, <\/strong>4]]<\/strong>
x_data<\/strong><\/a> = <\/strong>torch.tensor<\/strong><\/a>(data)<\/strong><\/p>\n\n\n\n

\u4ece NumPy \u6570\u7ec4\u521b\u5efa<\/strong><\/p>\n\n\n\n

\u5f20\u91cf\u53ef\u4ee5\u4ece NumPy \u6570\u7ec4\u521b\u5efa\uff08\u53cd\u4e4b\u4ea6\u7136 – \u8be6\u89c1\u4e0e NumPy \u7684\u6865\u63a5<\/a>\uff09\u3002<\/p>\n\n\n\n

np_array = np.array(data)<\/strong>
x_np<\/strong><\/a> = <\/strong>torch.from_numpy<\/strong><\/a>(np_array)<\/strong><\/p>\n\n\n\n\n\n\n\n

\u4ece\u53e6\u4e00\u4e2a\u5f20\u91cf\u521b\u5efa<\/strong><\/p>\n\n\n\n

\u65b0\u5f20\u91cf\u4f1a\u4fdd\u7559\u53c2\u6570\u5f20\u91cf\u7684\u5c5e\u6027\uff08\u5f62\u72b6\u3001\u6570\u636e\u7c7b\u578b\uff09\uff0c\u9664\u975e\u663e\u5f0f\u8986\u76d6\u3002<\/p>\n\n\n\n

x_ones<\/strong><\/a> = <\/strong>torch.ones_like<\/strong><\/a>(<\/strong>x_data<\/strong><\/a>) <\/strong># retains the properties of x_data<\/em>
print(<\/strong>f”Ones Tensor: \\n <\/strong>{<\/em>x_ones<\/strong><\/a>} <\/em>\\n<\/strong>“)<\/strong><\/p>\n\n\n\n

x_rand<\/strong><\/a> = <\/strong>torch.rand_like<\/strong><\/a>(<\/strong>x_data<\/strong><\/a>, dtype=<\/strong>torch.float<\/strong><\/a>) <\/strong># overrides the datatype of x_data<\/em>
print(<\/strong>f”Random Tensor: \\n <\/strong>{<\/em>x_rand<\/strong><\/a>} <\/em>\\n<\/strong>“)<\/strong><\/p>\n\n\n\n

Ones Tensor:
 tensor([[1, 1],
        [1, 1]])<\/p>\n\n\n\n

Random Tensor:
 tensor([[0.8823, 0.9150],
        [0.3829, 0.9593]])<\/p>\n\n\n\n

\u4f7f\u7528\u968f\u673a\u503c\u6216\u5e38\u91cf\u503c\u521b\u5efa<\/strong><\/p>\n\n\n\n

shape \u662f\u4e00\u4e2a\u5f20\u91cf\u7ef4\u5ea6\u7684\u5143\u7ec4\u3002\u5728\u4e0b\u9762\u7684\u51fd\u6570\u4e2d\uff0c\u5b83\u51b3\u5b9a\u4e86\u8f93\u51fa\u5f20\u91cf\u7684\u7ef4\u5ea6\u3002<\/p>\n\n\n\n

shape = (<\/strong>2,<\/strong>3,)<\/strong>
rand_tensor<\/strong><\/a> = <\/strong>torch.rand<\/strong><\/a>(shape)<\/strong>
ones_tensor<\/strong><\/a> = <\/strong>torch.ones<\/strong><\/a>(shape)<\/strong>
zeros_tensor<\/strong><\/a> = <\/strong>torch.zeros<\/strong><\/a>(shape)<\/strong><\/p>\n\n\n\n

print(<\/strong>f”Random Tensor: \\n <\/strong>{<\/em>rand_tensor<\/strong><\/a>} <\/em>\\n<\/strong>“)<\/strong>
print(<\/strong>f”Ones Tensor: \\n <\/strong>{<\/em>ones_tensor<\/strong><\/a>} <\/em>\\n<\/strong>“)<\/strong>
print(<\/strong>f”Zeros Tensor: \\n <\/strong>{<\/em>zeros_tensor<\/strong><\/a>}<\/em>“)<\/strong><\/p>\n\n\n\n

Random Tensor:
 tensor([[0.3904, 0.6009, 0.2566],
        [0.7936, 0.9408, 0.1332]])<\/p>\n\n\n\n

Ones Tensor:
 tensor([[1., 1., 1.],
        [1., 1., 1.]])<\/p>\n\n\n\n

Zeros Tensor:
 tensor([[0., 0., 0.],
        [0., 0., 0.]])<\/p>\n\n\n\n

\u5f20\u91cf\u7684\u5c5e\u6027<\/p>\n\n\n\n

\u5f20\u91cf\u5c5e\u6027\u63cf\u8ff0\u4e86\u5b83\u4eec\u7684\u5f62\u72b6\u3001\u6570\u636e\u7c7b\u578b\u4ee5\u53ca\u5b58\u50a8\u5b83\u4eec\u7684\u8bbe\u5907\u3002<\/p>\n\n\n\n

tensor<\/strong><\/a> = <\/strong>torch.rand<\/strong><\/a>(<\/strong>3,<\/strong>4)<\/strong><\/p>\n\n\n\n

print(<\/strong>f”Shape of tensor: {<\/em>tensor.shape<\/strong><\/a>}<\/em>“)<\/strong>
print(<\/strong>f”Datatype of tensor: {<\/em>tensor.dtype<\/strong><\/a>}<\/em>“)<\/strong>
print(<\/strong>f”Device tensor is stored on: {<\/em>tensor.device<\/strong><\/a>}<\/em>“)<\/strong><\/p>\n\n\n\n

Shape of tensor: torch.Size([3, 4])
Datatype of tensor: torch.float32
Device tensor is stored on: cpu<\/p>\n\n\n\n

\u5f20\u91cf\u64cd\u4f5c<\/p>\n\n\n\n

\u8d85\u8fc7 1200 \u79cd\u5f20\u91cf\u64cd\u4f5c\uff0c\u5305\u62ec\u7b97\u672f\u3001\u7ebf\u6027\u4ee3\u6570\u3001\u77e9\u9635\u64cd\u4f5c\uff08\u8f6c\u7f6e\u3001\u7d22\u5f15\u3001\u5207\u7247\uff09\u3001\u91c7\u6837\u7b49\u7b49\uff0c\u90fd\u5728\u6b64\u5904<\/a>\u8fdb\u884c\u4e86\u5168\u9762\u63cf\u8ff0\u3002<\/p>\n\n\n\n

\u8fd9\u4e9b\u64cd\u4f5c\u90fd\u53ef\u4ee5\u5728 CPU \u548c \u52a0\u901f\u5668<\/a>\u4e0a\u8fd0\u884c\uff0c\u4f8b\u5982 CUDA\u3001MPS\u3001MTIA \u6216 XPU\u3002\u5982\u679c\u60a8\u4f7f\u7528 Colab\uff0c\u53ef\u4ee5\u901a\u8fc7\u524d\u5f80\u201c\u8fd0\u884c\u65f6\u201d>\u201c\u66f4\u6539\u8fd0\u884c\u65f6\u7c7b\u578b\u201d>\u201cGPU\u201d\u6765\u5206\u914d\u4e00\u4e2a\u52a0\u901f\u5668\u3002<\/p>\n\n\n\n

\u9ed8\u8ba4\u60c5\u51b5\u4e0b\uff0c\u5f20\u91cf\u5728 CPU \u4e0a\u521b\u5efa\u3002\u6211\u4eec\u9700\u8981\u4f7f\u7528 .to \u65b9\u6cd5\uff08\u5728\u68c0\u67e5\u52a0\u901f\u5668\u53ef\u7528\u6027\u540e\uff09\u5c06\u5f20\u91cf\u663e\u5f0f\u79fb\u52a8\u5230\u52a0\u901f\u5668\u4e0a\u3002\u8bf7\u8bb0\u4f4f\uff0c\u5728\u8bbe\u5907\u4e4b\u95f4\u590d\u5236\u5927\u578b\u5f20\u91cf\u53ef\u80fd\u4f1a\u6d88\u8017\u5927\u91cf\u65f6\u95f4\u548c\u5185\u5b58\uff01<\/p>\n\n\n\n

# We move our tensor to the current accelerator if available<\/em>
if <\/strong>torch.accelerator.is_available<\/strong><\/a>():<\/strong>
    tensor<\/strong><\/a> <\/em>= <\/strong>tensor<\/strong><\/a>.to(<\/strong>torch.accelerator.current_accelerator<\/strong><\/a>())<\/strong><\/p>\n\n\n\n

\u5c1d\u8bd5\u5217\u8868\u4e2d\u7684\u4e00\u4e9b\u64cd\u4f5c\u3002\u5982\u679c\u60a8\u719f\u6089 NumPy API\uff0c\u60a8\u4f1a\u53d1\u73b0 Tensor API \u975e\u5e38\u6613\u4e8e\u4f7f\u7528\u3002<\/p>\n\n\n\n

\u6807\u51c6\u7684 NumPy \u5f0f\u7d22\u5f15\u548c\u5207\u7247<\/strong><\/p>\n\n\n\n

tensor<\/strong><\/a> = <\/strong>torch.ones<\/strong><\/a>(<\/strong>4, <\/strong>4)<\/strong>
print(<\/strong>f”First row: {<\/em>tensor<\/strong><\/a>[<\/strong>0]<\/strong>}<\/em>“)<\/strong>
print(<\/strong>f”First column: {<\/em>tensor<\/strong><\/a>[:, <\/strong>0]<\/strong>}<\/em>“)<\/strong>
print(<\/strong>f”Last column: {<\/em>tensor<\/strong><\/a>[…, –<\/strong>1]<\/strong>}<\/em>“)<\/strong>
tensor<\/strong><\/a>[:,<\/strong>1] = <\/strong>0
print(<\/strong>tensor<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

First row: tensor([1., 1., 1., 1.])
First column: tensor([1., 1., 1., 1.])
Last column: tensor([1., 1., 1., 1.])
tensor([[1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.]])<\/p>\n\n\n\n

\u8fde\u63a5\u5f20\u91cf<\/strong>\u00a0\u60a8\u53ef\u4ee5\u4f7f\u7528\u00a0torch.cat\u00a0\u6cbf\u7740\u7ed9\u5b9a\u7ef4\u5ea6\u8fde\u63a5\u4e00\u7cfb\u5217\u5f20\u91cf\uff08\u6cbf\u7740\u5750\u6807\u8f74\u62fc\u63a5<\/mark>\uff09\u3002\u53e6\u8bf7\u53c2\u9605\u00a0torch.stack<\/a>\uff0c\u8fd9\u662f\u53e6\u4e00\u4e2a\u5f20\u91cf\u8fde\u63a5\u64cd\u4f5c\uff0c\u4e0e\u00a0torch.cat\u00a0\u6709\u5fae\u5999\u7684\u533a\u522b\u3002<\/p>\n\n\n\n

t1<\/strong><\/a> = <\/strong>torch.cat<\/strong><\/a>([<\/strong>tensor<\/strong><\/a>, <\/strong>tensor<\/strong><\/a>, <\/strong>tensor<\/strong><\/a>], dim=<\/strong>1)<\/strong>
print(<\/strong>t1<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

tensor([[1., 0., 1., 1., 1., 0., 1., 1., 1., 0., 1., 1.],
        [1., 0., 1., 1., 1., 0., 1., 1., 1., 0., 1., 1.],
        [1., 0., 1., 1., 1., 0., 1., 1., 1., 0., 1., 1.],
        [1., 0., 1., 1., 1., 0., 1., 1., 1., 0., 1., 1.]])<\/p>\n\n\n\n

\u7b97\u672f\u8fd0\u7b97<\/strong><\/p>\n\n\n\n

# This computes the matrix multiplication between two tensors. y1, y2, y3 will have the same value<\/em>
# “tensor.T“ returns the transpose of a tensor  <\/em>@\u77e9\u9635\u4e58\u79ef<\/em> *<\/em>\u54c8\u8fbe\u739b\u79ef<\/em>\u5bf9\u5e94\u5143\u7d20\u76f8\u4e58<\/em>
y1<\/strong><\/a> <\/em>= <\/strong>tensor<\/strong><\/a> <\/em>@ <\/strong>tensor.T<\/strong><\/a>
y2<\/strong><\/a> <\/em>= <\/strong>tensor<\/strong><\/a>.matmul(<\/strong>tensor.T<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

y3<\/strong><\/a> = <\/strong>torch.rand_like<\/strong><\/a>(<\/strong>y1<\/strong><\/a>)<\/strong>
torch.matmul<\/strong><\/a>(<\/strong>tensor<\/strong><\/a>, <\/strong>tensor.T<\/strong><\/a>, out=<\/strong>y3<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

# This computes the element-wise product. z1, z2, z3 will have the same value<\/em>
z1<\/strong><\/a> <\/em>= <\/strong>tensor<\/strong><\/a> <\/em>* <\/strong>tensor<\/strong><\/a>
z2<\/strong><\/a> <\/em>= <\/strong>tensor<\/strong><\/a>.mul(<\/strong>tensor<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

z3<\/strong><\/a> = <\/strong>torch.rand_like<\/strong><\/a>(<\/strong>tensor<\/strong><\/a>)<\/strong>
torch.mul<\/strong><\/a>(<\/strong>tensor<\/strong><\/a>, <\/strong>tensor<\/strong><\/a>, out=<\/strong>z3<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

tensor([[1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.]])<\/p>\n\n\n\n

\u5355\u5143\u7d20\u5f20\u91cf<\/strong> \u5982\u679c\u60a8\u6709\u4e00\u4e2a\u5355\u5143\u7d20\u5f20\u91cf\uff0c\u4f8b\u5982\u901a\u8fc7\u5c06\u5f20\u91cf\u7684\u6240\u6709\u503c\u805a\u5408\u5230\u4e00\u4e2a\u503c\u4e2d\u83b7\u5f97\uff0c\u60a8\u53ef\u4ee5\u4f7f\u7528 item() \u5c06\u5176\u8f6c\u6362\u4e3a Python \u6570\u503c\u3002<\/p>\n\n\n\n

agg<\/strong><\/a> = <\/strong>tensor<\/strong><\/a>.sum()<\/strong>
agg_item = <\/strong>agg<\/strong><\/a>.item()<\/strong>
print(agg_item, <\/strong>type(agg_item))<\/strong><\/p>\n\n\n\n

12.0 <class ‘float’><\/p>\n\n\n\n

\u5c31\u5730\u64cd\u4f5c<\/strong> \u5c06\u7ed3\u679c\u5b58\u50a8\u5230\u64cd\u4f5c\u6570\u4e2d\u7684\u64cd\u4f5c\u79f0\u4e3a\u5c31\u5730\u64cd\u4f5c\u3002\u5b83\u4eec\u4ee5 _ \u540e\u7f00\u8868\u793a\u3002\u4f8b\u5982\uff1ax.copy_(y)\u3001x.t_() \u4f1a\u6539\u53d8 x\u3002<\/p>\n\n\n\n

print(<\/strong>f”{<\/em>tensor<\/strong><\/a>} <\/em>\\n<\/strong>“)<\/strong>
tensor<\/strong><\/a>.add_(<\/strong>5)<\/strong>
print(<\/strong>tensor<\/strong><\/a>)<\/strong><\/p>\n\n\n\n

tensor([[1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.],
        [1., 0., 1., 1.]])<\/p>\n\n\n\n

tensor([[6., 5., 6., 6.],
        [6., 5., 6., 6.],
        [6., 5., 6., 6.],
        [6., 5., 6., 6.]])<\/p>\n\n\n\n

\u6ce8\u610f<\/p>\n\n\n\n

\u5c31\u5730\u64cd\u4f5c\u8282\u7701\u4e86\u4e00\u4e9b\u5185\u5b58\uff0c\u4f46\u5728\u8ba1\u7b97\u5bfc\u6570\u65f6\u53ef\u80fd\u4f1a\u6709\u95ee\u9898\uff0c\u56e0\u4e3a\u4f1a\u7acb\u5373\u4e22\u5931\u5386\u53f2\u8bb0\u5f55\u3002\u56e0\u6b64\uff0c\u4e0d\u5efa\u8bae\u4f7f\u7528\u5b83\u4eec\u3002<\/p>\n\n\n\n

\u4e0e NumPy \u7684\u6865\u63a5<\/p>\n\n\n\n

CPU \u4e0a\u7684\u5f20\u91cf\u548c NumPy \u6570\u7ec4\u53ef\u4ee5\u5171\u4eab\u5176\u5e95\u5c42\u5185\u5b58\u4f4d\u7f6e\uff0c\u6539\u53d8\u5176\u4e2d\u4e00\u4e2a\u4f1a\u6539\u53d8\u53e6\u4e00\u4e2a\u3002<\/mark><\/p>\n\n\n\n

\u5f20\u91cf\u8f6c NumPy \u6570\u7ec4<\/p>\n\n\n\n

t<\/strong><\/a> = <\/strong>torch.ones<\/strong><\/a>(<\/strong>5)<\/strong>
print(<\/strong>f”t: {<\/em>t<\/strong><\/a>}<\/em>“)<\/strong>
n = <\/strong>t<\/strong><\/a>.numpy()<\/strong>
print(<\/strong>f”n: {<\/em>n<\/strong>}<\/em>“)<\/strong><\/p>\n\n\n\n

t: tensor([1., 1., 1., 1., 1.])
n: [1. 1. 1. 1. 1.]<\/p>\n\n\n\n

\u5f20\u91cf\u4e2d\u7684\u6539\u53d8\u4f1a\u53cd\u6620\u5728 NumPy \u6570\u7ec4\u4e2d\u3002<\/p>\n\n\n\n

t<\/strong><\/a>.add_(<\/strong>1)<\/strong>
print(<\/strong>f”t: {<\/em>t<\/strong><\/a>}<\/em>“)<\/strong>
print(<\/strong>f”n: {<\/em>n<\/strong>}<\/em>“)<\/strong><\/p>\n\n\n\n

t: tensor([2., 2., 2., 2., 2.])
n: [2. 2. 2. 2. 2.]<\/p>\n\n\n\n

NumPy \u6570\u7ec4\u8f6c\u5f20\u91cf<\/p>\n\n\n\n

n = np.ones(<\/strong>5)<\/strong>
t<\/strong><\/a> = <\/strong>torch.from_numpy<\/strong><\/a>(n)<\/strong><\/p>\n\n\n\n

NumPy \u6570\u7ec4\u4e2d\u7684\u6539\u53d8\u4f1a\u53cd\u6620\u5728\u5f20\u91cf\u4e2d\u3002<\/p>\n\n\n\n

np.add(n, <\/strong>1, out=n)<\/strong>
print(<\/strong>f”t: {<\/em>t<\/strong><\/a>}<\/em>“)<\/strong>
print(<\/strong>f”n: {<\/em>n<\/strong>}<\/em>“)<\/strong><\/p>\n\n\n\n

t: tensor([2., 2., 2., 2., 2.], dtype=torch.float64)
n: [2. 2. 2. 2. 2.]<\/p>\n\n\n\n

\u5982\u679c\u5c06tensor\u79fb\u52a8\u5230GPU\uff0c\u548cNumPy\u4e0d\u518d\u5171\u4eab<\/mark><\/p>\n\n\n\n

t = torch.ones(5)
print(f”t: {t}”)
n = t.numpy()
print(f”n: {n}”)
t.add_(1)
print(f”t: {t}”)
print(f”n: {n}”)<\/p>\n\n\n\n

if torch.accelerator.is_available():
    t = t.to(torch.accelerator.current_accelerator())
t.add_(1)
print(f”t: {t}”)
print(f”n: {n}”)<\/p>\n\n\n\n

t: tensor([1., 1., 1., 1., 1.])
n: [1. 1. 1. 1. 1.]
t: tensor([2., 2., 2., 2., 2.])
n: [2. 2. 2. 2. 2.]
t: tensor([3., 3., 3., 3., 3.], device=’mps:0′)
n: [2. 2. 2. 2. 2.]<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5f20\u91cf\u662f\u4e00\u79cd\u7279\u6b8a\u7684\u6570\u636e\u7ed3\u6784\uff0c\u4e0e\u6570\u7ec4\u548c\u77e9\u9635\u975e\u5e38\u76f8\u4f3c\u3002\u5728 PyTorch \u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u5f20\u91cf\u6765\u7f16\u7801\u6a21\u578b\u7684\u8f93\u5165\u548c\u8f93\u51fa\uff0c\u4ee5 … <\/p>\n

\u7ee7\u7eed\u9605\u8bfb\u201ctorch \u5f20\u91cf\u201d<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[246,14],"tags":[9,254,253],"class_list":["post-2471","post","type-post","status-publish","format-standard","hentry","category-ai","category-python","tag-python","tag-tensor","tag-torch"],"_links":{"self":[{"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/posts\/2471","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/thereisno.top\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2471"}],"version-history":[{"count":1,"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/posts\/2471\/revisions"}],"predecessor-version":[{"id":2472,"href":"https:\/\/thereisno.top\/index.php?rest_route=\/wp\/v2\/posts\/2471\/revisions\/2472"}],"wp:attachment":[{"href":"https:\/\/thereisno.top\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2471"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/thereisno.top\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2471"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/thereisno.top\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2471"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}