Merge pull request #1 from calincru/master

Use a conda .environment file to easily set up the dev environment and reproduce the behaviour

.environment

@@ -0,0 +1,16 @@
+dependencies:
+  - python=3.5
+  - numpy=1.14
+  - scipy=1.0
+  - matplotlib=2.2
+  - pyzmq=17.0
+  - jupyter=1.0
+  - cython=0.28
+  - pytorch=0.3  # from channel pytorch
+  - torchvision=0.2  # from channel pytorch
+  - shogun=6.0  # from channel conda-forge
+  - tqdm
+  - pylint
+  - sphinxcontrib-bibtex
+  - sphinx_rtd_theme
+  - mock

.gitignore

@@ -12,3 +12,4 @@ build
 *.swo
 dist
 docs/_build
+.pytest_cache/

.travis.yml

@@ -19,16 +19,14 @@ install:
   - hash -r
   - conda config --set always_yes yes --set changeps1 no
   - conda update -q conda
+  # Add channels necessary for the dependencies
+  - conda config --add channels pytorch
+  - conda config --add channels conda-forge
   # Useful for debugging any issues with conda
   - conda info -a
 
-  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION
+  - conda env create -n test-environment -f .environment python=$TRAVIS_PYTHON_VERSION
   - source activate test-environment
-  - conda install pytorch -c soumith
-  - conda install -c conda-forge shogun 
-  - conda install numpy
-  - conda install cython
-  - conda install scipy
   - python setup.py install
 
 script:

notebooks/.gitignore

@@ -0,0 +1,2 @@
+.ipynb_checkpoints/
+mnist_dir/

tests/inference_cardinality_test.py

@@ -14,7 +14,7 @@ def test_softmax():
             unaries = Variable(torch.randn(batch_size, 15))
             count_potentials = Variable(NINF * torch.ones(batch_size, k + 1))
             count_potentials[:, 1] = 0
-            output_softmax = softmax(unaries)
+            output_softmax = softmax(unaries, dim=1)
             output_cardinf = inference_cardinality(unaries, count_potentials)
             assert np.allclose(output_softmax.cpu().data.numpy(),
                                output_cardinf.cpu().data.numpy())

torch_two_sample/inference_cardinality.py

@@ -139,8 +139,9 @@ def create_var(val, *dims):
 
     bb = bmsgs[:, 2:, :]
     ff = fmsgs[:, :-2, :]
-    b0 = logsumexp(bb + ff, 2)
-    b1 = logsumexp(bb[:, :, :-1] + ff[:, :, 1:], 2) + node_potentials[:, :-1]
+    b0 = logsumexp(bb + ff, 2).view(batch_size, dim_node-1)
+    b1 = logsumexp(bb[:, :, :-1] + ff[:, :, 1:], 2).view(
+        batch_size, dim_node-1) + node_potentials[:, :-1]
 
     marginals = create_var(0, batch_size, dim_node)
     marginals[:, :-1] = torch.sigmoid(b1 - b0)

torch_two_sample/statistics_diff.py

@@ -244,7 +244,7 @@ def __call__(self, sample_1, sample_2, alphas, norm=2, ret_matrix=False):
         margs_ = None
         for alpha in alphas:
             if self.k == 1:
-                margs_a = softmax(- alpha * diffs)
+                margs_a = softmax(-alpha * diffs, dim=1)
             else:
                 margs_a = inference_cardinality(
                     - alpha * diffs.cpu(), count_potential)