Refactor into proper package

data/bsds300.py

@@ -1,7 +1,7 @@
 import h5py
 import numpy as np
 import os
-import utils
+from nsflow import utils
 
 from matplotlib import pyplot as plt
 from torch.utils import data

data/gas.py

@@ -1,7 +1,7 @@
 import numpy as np
 import os
 import pandas as pd
-import utils
+from nsflow import utils
 
 from matplotlib import pyplot as plt
 from torch.utils.data import Dataset

data/hepmass.py

@@ -1,7 +1,7 @@
 import numpy as np
 import os
 import pandas as pd
-import utils
+from nsflow import utils
 
 from collections import Counter
 from matplotlib import pyplot as plt

data/miniboone.py

@@ -1,6 +1,6 @@
 import numpy as np
 import os
-import utils
+from nsflow import utils
 
 from matplotlib import pyplot as plt
 from torch.utils.data import Dataset

data/omniglot.py

@@ -2,7 +2,7 @@
 import os
 import torch
 
-import utils
+from nsflow import utils
 
 from PIL import Image
 from scipy.io import loadmat

data/plane.py

@@ -8,7 +8,7 @@
 from torch import distributions
 from torch.utils.data import Dataset
 
-import utils
+from nsflow import utils
 
 
 class PlaneDataset(Dataset):

data/power.py

@@ -1,6 +1,6 @@
 import numpy as np
 import os
-import utils
+from nsflow import utils
 
 from matplotlib import pyplot as plt
 from torch.utils.data import Dataset

nde/distributions/base.py → nsflow/nde/distributions/base.py

@@ -4,7 +4,7 @@
 
 from torch import nn
 
-import utils
+from nsflow import utils
 
 
 class NoMeanException(Exception):

nde/distributions/discrete.py → nsflow/nde/distributions/discrete.py

@@ -4,9 +4,9 @@
 
 from torch.nn import functional as F
 
-import utils
+from nsflow import utils
 
-from nde import distributions
+from nsflow.nde import distributions
 
 
 class ConditionalIndependentBernoulli(distributions.Distribution):

nde/distributions/normal.py → nsflow/nde/distributions/normal.py

@@ -3,9 +3,9 @@
 import numpy as np
 import torch
 
-import utils
+from nsflow import utils
 
-from nde import distributions
+from nsflow.nde import distributions
 
 
 class StandardNormal(distributions.Distribution):

nde/distributions/discrete_test.py → nsflow/nde/distributions/tests/test_discrete.py

@@ -3,7 +3,7 @@
 import torch
 import torchtestcase
 import unittest
-from nde.distributions import discrete
+from nsflow.nde.distributions import discrete
 
 
 class ConditionalIndependentBernoulliTest(torchtestcase.TorchTestCase):

nde/distributions/normal_test.py → nsflow/nde/distributions/tests/test_normal.py

@@ -3,7 +3,7 @@
 import torch
 import torchtestcase
 import unittest
-from nde.distributions import normal
+from nsflow.nde.distributions import normal
 
 
 class StandardNormalTest(torchtestcase.TorchTestCase):

nde/flows/autoregressive.py → nsflow/nde/flows/autoregressive.py

@@ -2,9 +2,9 @@
 
 from torch.nn import functional as F
 
-from nde import distributions
-from nde import flows
-from nde import transforms
+from nsflow.nde import distributions
+from nsflow.nde import flows
+from nsflow.nde import transforms
 
 
 class MaskedAutoregressiveFlow(flows.Flow):

nde/flows/base.py → nsflow/nde/flows/base.py

@@ -1,8 +1,8 @@
 """Basic definitions for the flows module."""
 
-import utils
+from nsflow import utils
 
-from nde import distributions
+from nsflow.nde import distributions
 
 
 class Flow(distributions.Distribution):

nde/flows/realnvp.py → nsflow/nde/flows/realnvp.py

@@ -3,10 +3,10 @@
 import torch
 from torch.nn import functional as F
 
-from nde import distributions
-from nde import flows
-from nde import transforms
-import nn as nn_
+from nsflow.nde import distributions
+from nsflow.nde import flows
+from nsflow.nde import transforms
+from nsflow import nn as nn_
 
 
 class SimpleRealNVP(flows.Flow):

nde/flows/autoregressive_test.py → nsflow/nde/flows/tests/test_autoregressive.py

@@ -4,7 +4,7 @@
 import torchtestcase
 import unittest
 
-from nde.flows import autoregressive as ar
+from nsflow.nde.flows import autoregressive as ar
 
 
 class MaskedAutoregressiveFlowTest(torchtestcase.TorchTestCase):

nde/flows/base_test.py → nsflow/nde/flows/tests/test_base.py

@@ -3,9 +3,9 @@
 import torch
 import torchtestcase
 import unittest
-from nde import transforms
-from nde import distributions
-from nde.flows import base
+from nsflow.nde import transforms
+from nsflow.nde import distributions
+from nsflow.nde.flows import base
 
 
 class FlowTest(torchtestcase.TorchTestCase):

nde/flows/realnvp_test.py → nsflow/nde/flows/tests/test_realnvp.py

@@ -4,7 +4,7 @@
 import torchtestcase
 import unittest
 
-from nde.flows import realnvp
+from nsflow.nde.flows import realnvp
 
 
 class SimpleRealNVPTest(torchtestcase.TorchTestCase):

nde/transforms/autoregressive.py → nsflow/nde/transforms/autoregressive.py

@@ -4,10 +4,10 @@
 import torch
 from torch.nn import functional as F
 
-import utils
-from nde import transforms
-from nde.transforms import made as made_module
-from nde.transforms import splines
+from nsflow import utils
+from nsflow.nde import transforms
+from nsflow.nde.transforms import made as made_module
+from nsflow.nde.transforms import splines
 
 class AutoregressiveTransform(transforms.Transform):
     """Transforms each input variable with an invertible elementwise transformation.

nde/transforms/base.py → nsflow/nde/transforms/base.py

@@ -5,7 +5,7 @@
 
 from torch import nn
 
-import utils
+from nsflow import utils
 
 
 class InverseNotAvailable(Exception):

nde/transforms/conv.py → nsflow/nde/transforms/conv.py

@@ -1,5 +1,6 @@
-import utils
-from nde import transforms
+from nsflow import utils
+
+from nsflow.nde import transforms
 
 
 class OneByOneConvolution(transforms.LULinear):

nde/transforms/coupling.py → nsflow/nde/transforms/coupling.py

@@ -5,10 +5,10 @@
 import torch
 from torch.nn import functional as F
 
-import utils
+from nsflow import utils
 
-from nde import transforms
-from nde.transforms import splines
+from nsflow.nde import transforms
+from nsflow.nde.transforms import splines
 
 
 class CouplingTransform(transforms.Transform):

nde/transforms/linear.py → nsflow/nde/transforms/linear.py

@@ -6,9 +6,9 @@
 from torch import nn
 from torch.nn import functional as F, init
 
-import utils
+from nsflow import utils
 
-from nde import transforms
+from nsflow.nde import transforms
 
 
 class LinearCache(object):
@@ -177,7 +177,7 @@ def inverse_no_cache(self, inputs):
         """
         batch_size = inputs.shape[0]
         outputs = inputs - self.bias
-        outputs, lu = torch.gesv(outputs.t(), self._weight)  # Linear-system solver.
+        outputs, lu = torch.solve(outputs.t(), self._weight)  # Linear-system solver.
         outputs = outputs.t()
         # The linear-system solver returns the LU decomposition of the weights, which we
         # can use to obtain the log absolute determinant directly.
@@ -210,7 +210,7 @@ def weight_inverse_and_logabsdet(self):
         """
         # If both weight inverse and logabsdet are needed, it's cheaper to compute both together.
         identity = torch.eye(self.features, self.features)
-        weight_inv, lu = torch.gesv(identity, self._weight)  # Linear-system solver.
+        weight_inv, lu = torch.solve(identity, self._weight)  # Linear-system solver.
         logabsdet = torch.sum(torch.log(torch.abs(torch.diag(lu))))
         return weight_inv, logabsdet
 

nde/transforms/lu.py → nsflow/nde/transforms/lu.py

@@ -4,7 +4,7 @@
 from torch import nn
 from torch.nn import functional as F, init
 
-from nde.transforms.linear import Linear
+from nsflow.nde.transforms.linear import Linear
 
 
 class LULinear(Linear):
@@ -103,8 +103,8 @@ def weight_inverse(self):
         """
         lower, upper = self._create_lower_upper()
         identity = torch.eye(self.features, self.features)
-        lower_inverse, _ = torch.trtrs(identity, lower, upper=False, unitriangular=True)
-        weight_inverse, _ = torch.trtrs(lower_inverse, upper, upper=True, unitriangular=False)
+        lower_inverse, _ = torch.triangular_solve(identity, lower, upper=False, unitriangular=True)
+        weight_inverse, _ = torch.triangular_solve(lower_inverse, upper, upper=True, unitriangular=False)
         return weight_inverse
 
     @property
@@ -117,4 +117,4 @@ def logabsdet(self):
         where:
             D = num of features
         """
-        return torch.sum(torch.log(self.upper_diag))
+        return torch.sum(torch.log(self.upper_diag))

nde/transforms/made.py → nsflow/nde/transforms/made.py

@@ -3,7 +3,7 @@
 
 import torch
 
-import utils
+from nsflow import utils
 
 from torch import nn
 from torch.nn import functional as F, init

nde/transforms/nonlinearities.py → nsflow/nde/transforms/nonlinearities.py

@@ -6,10 +6,10 @@
 from torch import nn
 from torch.nn import functional as F
 
-import utils
+from nsflow import utils
 
-from nde import transforms
-from nde.transforms import splines
+from nsflow.nde import transforms
+from nsflow.nde.transforms import splines
 
 class Tanh(transforms.Transform):
     def forward(self, inputs, context=None):

nde/transforms/normalization.py → nsflow/nde/transforms/normalization.py

@@ -5,9 +5,9 @@
 from torch import nn
 from torch.nn import functional as F
 
-import utils
+from nsflow import utils
 
-from nde import transforms
+from nsflow.nde import transforms
 
 
 # class BatchNorm(transforms.Transform):

nde/transforms/orthogonal.py → nsflow/nde/transforms/orthogonal.py

@@ -4,9 +4,9 @@
 
 from torch import nn
 
-import utils
+from nsflow import utils
 
-from nde import transforms
+from nsflow.nde import transforms
 
 
 class HouseholderSequence(transforms.Transform):

nde/transforms/permutations.py → nsflow/nde/transforms/permutations.py

@@ -1,9 +1,9 @@
 """Implementations of permutation-like transforms."""
 
 import torch
-import utils
+from nsflow import utils
 
-from nde import transforms
+from nsflow.nde import transforms
 
 
 class Permutation(transforms.Transform):

nde/transforms/qr.py → nsflow/nde/transforms/qr.py

@@ -4,8 +4,8 @@
 from torch import nn
 from torch.nn import functional as F, init
 
-from nde import transforms
-from nde.transforms.linear import Linear
+from nsflow.nde import transforms
+from nsflow.nde.transforms.linear import Linear
 
 
 class QRLinear(Linear):
@@ -70,7 +70,7 @@ def inverse_no_cache(self, inputs):
         upper = self._create_upper()
         outputs = inputs - self.bias
         outputs, _ = self.orthogonal.inverse(outputs)  # Ignore logabsdet since we know it's zero.
-        outputs, _ = torch.trtrs(outputs.t(), upper, upper=True)
+        outputs, _ = torch.triangular_solve(outputs.t(), upper, upper=True)
         outputs = outputs.t()
         logabsdet = -self.logabsdet()
         logabsdet = logabsdet * torch.ones(outputs.shape[0])
@@ -96,7 +96,7 @@ def weight_inverse(self):
         """
         upper = self._create_upper()
         identity = torch.eye(self.features, self.features)
-        upper_inv, _ = torch.trtrs(identity, upper, upper=True)
+        upper_inv, _ = torch.triangular_solve(identity, upper, upper=True)
         weight_inv, _ = self.orthogonal(upper_inv)
         return weight_inv
 

nde/transforms/reshape.py → nsflow/nde/transforms/reshape.py

@@ -1,7 +1,8 @@
 import torch
 
-import utils
-from nde import transforms
+from nsflow import utils
+
+from nsflow.nde import transforms
 
 
 class SqueezeTransform(transforms.Transform):

nde/transforms/splines/cubic.py → nsflow/nde/transforms/splines/cubic.py

@@ -3,8 +3,8 @@
 import torch
 from torch.nn import functional as F
 
-import utils
-from nde import transforms
+from nsflow import utils
+from nsflow.nde import transforms
 
 DEFAULT_MIN_BIN_WIDTH = 1e-3
 DEFAULT_MIN_BIN_HEIGHT = 1e-3