torch.cummax#

torch.cummax(input,dim,*,out=None)#

Returns a namedtuple(values,indices) wherevalues is the cumulative maximum ofelements ofinput in the dimensiondim. Andindices is the indexlocation of each maximum value found in the dimensiondim.

$$y_i=max(x_1,x_2,x_3,\dots ,x_i)$$yi​=max(x1​,x2​,x3​,…,xi​)

Parameters

• input (Tensor) – the input tensor.

• dim (int) – the dimension to do the operation over

Keyword Arguments

out (tuple,optional) – the result tuple of two output tensors (values, indices)

Example:

>>>a=torch.randn(10)>>>atensor([-0.3449, -1.5447,  0.0685, -1.5104, -1.1706,  0.2259,  1.4696, -1.3284,     1.9946, -0.8209])>>>torch.cummax(a,dim=0)torch.return_types.cummax(    values=tensor([-0.3449, -0.3449,  0.0685,  0.0685,  0.0685,  0.2259,  1.4696,  1.4696,     1.9946,  1.9946]),    indices=tensor([0, 0, 2, 2, 2, 5, 6, 6, 8, 8]))