lovely-tensors 0.0.4

Lovely Tensors

Lovely Tensors

Install

pip install lovely-tensors

How to use

How often do you find yourself debuggin a neural network? You dump a tensor to the cell output, and see this:

numbers

tensor([[[-0.3541, -0.3369, -0.4054,  ..., -0.5596, -0.4739,  2.2489],
         [-0.4054, -0.4226, -0.4911,  ..., -0.9192, -0.8507,  2.1633],
         [-0.4739, -0.4739, -0.5424,  ..., -1.0390, -1.0390,  2.1975],
         ...,
         [-0.9020, -0.8335, -0.9363,  ..., -1.4672, -1.2959,  2.2318],
         [-0.8507, -0.7822, -0.9363,  ..., -1.6042, -1.5014,  2.1804],
         [-0.8335, -0.8164, -0.9705,  ..., -1.6555, -1.5528,  2.1119]],

        [[-0.1975, -0.1975, -0.3025,  ..., -0.4776, -0.3725,  2.4111],
         [-0.2500, -0.2325, -0.3375,  ..., -0.7052, -0.6702,  2.3585],
         [-0.3025, -0.2850, -0.3901,  ..., -0.7402, -0.8102,  2.3761],
         ...,
         [-0.4251, -0.2325, -0.3725,  ..., -1.0903, -1.0203,  2.4286],
         [-0.3901, -0.2325, -0.4251,  ..., -1.2304, -1.2304,  2.4111],
         [-0.4076, -0.2850, -0.4776,  ..., -1.2829, -1.2829,  2.3410]],

        [[-0.6715, -0.9853, -0.8807,  ..., -0.9678, -0.6890,  2.3960],
         [-0.7238, -1.0724, -0.9678,  ..., -1.2467, -1.0201,  2.3263],
         [-0.8284, -1.1247, -1.0201,  ..., -1.2641, -1.1596,  2.3786],
         ...,
         [-1.2293, -1.4733, -1.3861,  ..., -1.5081, -1.2641,  2.5180],
         [-1.1944, -1.4559, -1.4210,  ..., -1.6476, -1.4733,  2.4308],
         [-1.2293, -1.5256, -1.5081,  ..., -1.6824, -1.5256,  2.3611]]])

Was it really useful?

What is the shape?
What are the statistics?
Are any of the values nan or inf?
Is it an image of a man holding a tench?

import lovely_tensors.tensors as lt

lt.PRINT_OPTS.color=True

# A very short tensor - no min/max
print(lt.lovely(numbers.view(-1)[:2]))
# A slightly longer tensor
print(lt.lovely(numbers.view(-1)[:6].view(2,3)))

tensor[2] μ=-0.345 σ=0.012 x=[-0.354, -0.337]
tensor[2, 3] n=6 x∈[-0.440, -0.337] μ=-0.388 σ=0.038 x=[[-0.354, -0.337, -0.405], [-0.440, -0.388, -0.405]]

lt.lovely(numbers)

tensor[3, 196, 196] n=115248 x∈[-2.118, 2.640] μ=-0.388 σ=1.073

lt.lovely(numbers, depth=1)

tensor[3, 196, 196] n=115248 x∈[-2.118, 2.640] μ=-0.388 σ=1.073
  tensor[196, 196] n=38416 x∈[-2.118, 2.249] μ=-0.324 σ=1.036
  tensor[196, 196] n=38416 x∈[-1.966, 2.429] μ=-0.274 σ=0.973
  tensor[196, 196] n=38416 x∈[-1.804, 2.640] μ=-0.567 σ=1.178

t = numbers.view(-1)[:12].clone()

t[0] *= 10000
t[1] /= 10000
t[2] = float('inf')
t[3] = float('-inf')
t[4] = float('nan')
t = t.reshape((2,6))

# A spicy tensor
lt.lovely(t)

tensor[2, 6] n=12 x∈[-3.541e+03, -3.369e-05] μ=-393.776 σ=1.180e+03 +inf! -inf! nan!

# A zero tensor
lt.lovely(torch.zeros(10, 10))

tensor[10, 10] n=100 all_zeros

Now the important queston - is it our man?

lt.rgb(numbers)

Maaaaybe? Looks like someone normalized him.

in_stats = { "mean": (0.485, 0.456, 0.406),
             "std": (0.229, 0.224, 0.225) }
lt.rgb(numbers, in_stats)

There can be no doubt, it’s out hero the Tenchman!

One last thing - let’s monkey-patch torch.Tensor for convenience.

lt.monkey_patch()
t

tensor[2, 6] n=12 x∈[-3.541e+03, -3.369e-05] μ=-393.776 σ=1.180e+03 +inf! -inf! nan!

t.verbose

tensor[2, 6] n=12 x∈[-3.541e+03, -3.369e-05] μ=-393.776 σ=1.180e+03 +inf! -inf! nan!
x=[[-3.5405e+03, -3.3693e-05,         inf,        -inf,         nan, -4.0543e-01],
   [-4.2255e-01, -4.9105e-01, -5.0818e-01, -5.5955e-01, -5.4243e-01, -5.0818e-01]]

t.plain

[[-3.5405e+03, -3.3693e-05,         inf,        -inf,         nan, -4.0543e-01],
 [-4.2255e-01, -4.9105e-01, -5.0818e-01, -5.5955e-01, -5.4243e-01, -5.0818e-01]]

numbers.rgb
# you can also do numbers.rgb()

#per-channel stats
numbers.deeper

tensor[3, 196, 196] n=115248 x∈[-2.118, 2.640] μ=-0.388 σ=1.073
  tensor[196, 196] n=38416 x∈[-2.118, 2.249] μ=-0.324 σ=1.036
  tensor[196, 196] n=38416 x∈[-1.966, 2.429] μ=-0.274 σ=0.973
  tensor[196, 196] n=38416 x∈[-1.804, 2.640] μ=-0.567 σ=1.178

# You can go even deeper if you want to
dt = torch.randn(3, 3, 5)
dt.deeper(2)

tensor[3, 3, 5] n=45 x∈[-2.541, 2.153] μ=0.110 σ=0.988
  tensor[3, 5] n=15 x∈[-1.115, 2.153] μ=0.161 σ=0.967
    tensor[5] x∈[-1.115, 1.054] μ=0.217 σ=0.828 x=[0.612, 1.054, 0.501, 0.033, -1.115]
    tensor[5] x∈[-0.765, 2.153] μ=0.544 σ=1.264 x=[-0.765, 0.750, -0.688, 2.153, 1.271]
    tensor[5] x∈[-1.057, 0.930] μ=-0.277 σ=0.747 x=[-0.675, -0.239, 0.930, -1.057, -0.345]
  tensor[3, 5] n=15 x∈[-2.541, 2.136] μ=-0.056 σ=1.276
    tensor[5] x∈[-1.441, 1.657] μ=0.163 σ=1.137 x=[-1.441, -0.255, 0.584, 0.269, 1.657]
    tensor[5] x∈[-2.541, 2.136] μ=0.312 σ=1.894 x=[2.136, -0.509, 0.737, 1.735, -2.541]
    tensor[5] x∈[-1.273, -0.266] μ=-0.643 σ=0.400 x=[-0.646, -1.273, -0.331, -0.266, -0.700]
  tensor[3, 5] n=15 x∈[-1.279, 1.296] μ=0.225 σ=0.677
    tensor[5] x∈[-0.400, 1.296] μ=0.513 σ=0.614 x=[-0.400, 0.412, 0.738, 0.520, 1.296]
    tensor[5] x∈[-0.833, 0.788] μ=0.093 σ=0.605 x=[0.267, 0.788, -0.094, 0.339, -0.833]
    tensor[5] x∈[-1.279, 0.905] μ=0.069 σ=0.839 x=[-0.054, 0.905, 0.592, -1.279, 0.181]

# A quick de-norm. Don't worry, the data stays the same.
numbers.rgb(in_stats)