tensor.max, tensor.argmax

一个batch_size=32模型训练二分类预测输出的结果

output=torch.tensor([[0.4386, 0.6951],
        [0.8676, 0.8337],
        [0.5899, 0.5169],
        [0.4655, 0.7301],
        [0.5195, 0.8004],
        [0.0000, 0.0000],
        [0.9471, 0.9369],
        [0.5637, 0.8579],
        [0.7036, 0.4013],
        [0.0772, 0.0000],
        [0.4833, 0.7533],
        [0.5751, 0.8727],
        [0.7379, 0.6507],
        [0.5818, 0.5124],
        [0.5079, 0.7853],
        [0.3538, 0.5229],
        [0.8047, 0.9168],
        [0.5269, 0.8100],
        [0.4793, 0.6636],
        [0.5675, 0.5823],
        [0.0289, 0.0000],
        [0.5372, 0.7232],
        [0.5348, 0.8203],
        [0.5092, 0.5612],
        [0.5882, 0.7212],
        [0.5012, 0.7163],
        [0.6365, 0.8210],
        [0.0000, 0.0475],
        [0.5286, 0.7455],
        [0.5431, 0.8311],
        [0.7555, 0.8213],
        [0.6197, 0.8221]])

output.argmax(dim=-1)

每一行的最大值下标
tensor([1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1,
        1, 1, 1, 1, 1, 1, 1, 1])

output.argmax(dim=0)

在列上最大值的下标
tensor([6, 6])

aaa=torch.tensor([[1.8742, 0.0000],
        [2.8360, 0.0000],
        [2.2612, 0.0000],
        [2.0899, 0.0000],
        [1.4432, 0.0000],
        [2.1185, 0.0000],
        [1.9332, 0.0000],
        [1.9641, 0.0000],
        [2.1104, 0.0000],
        [1.8978, 0.0000],
        [1.7399, 0.0000],
        [0.7568, 0.0000],
        [2.1608, 0.0000],
        [2.4813, 0.0000],
        [1.7124, 0.0000],
        [1.7205, 0.0000],
        [0.8863, 0.0000],
        [0.5090, 0.0000],
        [1.3449, 0.0000],
        [0.4651, 0.0000],
        [1.7381, 0.0000],
        [0.4817, 0.0000],
        [2.4699, 0.0000],
        [1.9334, 0.0000],
        [2.2564, 0.0000],
        [0.4286, 0.0000],
        [2.1439, 0.0000],
        [0.2625, 0.0000],
        [1.4068, 0.0000],
        [2.1745, 0.0000],
        [1.7621, 0.0000],
        [1.5503, 100]])

aaa.argmax(dim=0)

在列上最大值下标
tensor([ 1, 31])

aaa.argmax(dim=-1)

在行上最大值下标
tensor([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 1])

max

aaa=torch.tensor([[1.8742, 0.0000],
        [2.8360, 0.0000],
        [2.2612, 0.0000],
        [2.0899, 0.0000],
        [1.4432, 0.0000],
        [2.1185, 0.0000],
        [1.9332, 0.0000],
        [1.9641, 0.0000],
        [2.1104, 0.0000],
        [1.8978, 0.0000],
        [1.7399, 0.0000],
        [0.7568, 0.0000],
        [2.1608, 0.0000],
        [2.4813, 0.0000],
        [1.7124, 0.0000],
        [1.7205, 0.0000],
        [0.8863, 0.0000],
        [0.5090, 0.0000],
        [1.3449, 0.0000],
        [0.4651, 0.0000],
        [1.7381, 0.0000],
        [0.4817, 0.0000],
        [2.4699, 0.0000],
        [1.9334, 0.0000],
        [2.2564, 0.0000],
        [0.4286, 0.0000],
        [2.1439, 0.0000],
        [0.2625, 0.0000],
        [1.4068, 0.0000],
        [2.1745, 0.0000],
        [1.7621, 0.0000],
        [1.5503, 100]])

aaa.max(dim=0)

取出在列上最大值及其下标
torch.return_types.max(
values=tensor([  2.8360, 100.0000]),
indices=tensor([ 1, 31]))

aaa.max(dim=1)

torch.return_types.max(
values=tensor([  1.8742,   2.8360,   2.2612,   2.0899,   1.4432,   2.1185,   1.9332,
          1.9641,   2.1104,   1.8978,   1.7399,   0.7568,   2.1608,   2.4813,
          1.7124,   1.7205,   0.8863,   0.5090,   1.3449,   0.4651,   1.7381,
          0.4817,   2.4699,   1.9334,   2.2564,   0.4286,   2.1439,   0.2625,
          1.4068,   2.1745,   1.7621, 100.0000]),
indices=tensor([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 1]))