pytorch代码

模型加载，从cpu到gpu

self.net.load_state_dict(torch.load(net_path,map_location=lambda storage,loc:storage))

namedetuple给元组起名字

namedtuple('GenericDict',cfg.keys())

np.hanning()

np.hanning(N)
#长为n的汉宁窗

np.outer()

a = np.arange(4)
array([0, 1, 2, 3])

np.outer(a,(2,1))
array([[0, 0],
       [2, 1],
       [4, 2],
       [6, 3]])

np.outer(a,1)
array([[0],
       [1],
       [2],
       [3]])

torch:
unsqueeze()在指定维度上加一维
squeeze()在指定维度上减一维
np:
title(a,())在指定维度上复制多次

a.shape
Out[43]: (4,)

torch.from_numpy(a).unsqueeze(1).shape
Out[41]: torch.Size([4, 1])

torch.from_numpy(a).unsqueeze(0).shape
Out[42]: torch.Size([1, 4])

torch.from_numpy(a).squeeze(0).shape
Out[44]: torch.Size([4])

np.tile(a,(2,1)).shape
Out[52]: (2, 4)

np.tile(a,(3,2,1)).shape
Out[51]: (3, 2, 4)