added codecov

justanhduc · Aug 17, 2020 · 6bda19f · 6bda19f
1 parent ed5c9fd
commit 6bda19f
Show file tree

Hide file tree

Showing 4 changed files with 65 additions and 67 deletions.
diff --git a/.travis.yml b/.travis.yml
@@ -21,7 +21,8 @@ install:
   - ./.travis/travis_install.sh
   - source activate pyenv
   - conda install -y pytorch torchvision cpuonly -c pytorch
-  - pip install pytest
+  - pip install pytest-cov pytest
+  - pip install codecov
   - pip install flake8
   - pip install -r requirements.txt
   - pip install .
@@ -44,8 +45,14 @@ script:
   - echo "$PART"
   - python -c 'import torch; print(torch.__version__)'
   - python -c 'import neuralnet_pytorch; print(neuralnet_pytorch.__version__)'
-  - python -m pytest -sv tests/
+  - python -m pytest -v --cov=neuralnet_pytorch/
   - flake8 neuralnet_pytorch
 
+env:
+  - CODECOV_TOKEN="41bdb269-8e93-42a6-a87d-65339e8145e9"
+
 after_failure:
-- cat /home/travis/.pip/pip.log
+- cat /home/travis/.pip/pip.log
+
+after_success:
+  - bash <(curl -s https://codecov.io/bash)
diff --git a/README.md b/README.md
@@ -6,6 +6,7 @@
 [![License](https://img.shields.io/github/license/justanhduc/neuralnet-pytorch.svg)](https://github.com/justanhduc/neuralnet-pytorch/blob/master/LICENSE.txt)
 [![Build Status](https://travis-ci.org/justanhduc/neuralnet-pytorch.svg?branch=master)](https://travis-ci.org/justanhduc/neuralnet-pytorch)
 [![Documentation Status](https://readthedocs.org/projects/neuralnet-pytorch/badge/?version=latest)](https://neuralnet-pytorch.readthedocs.io/en/latest/?badge=latest)
+[![codecov](https://codecov.io/gh/justanhduc/neuralnet-pytorch/branch/master/graph/badge.svg)](https://codecov.io/gh/justanhduc/neuralnet-pytorch)
 
 __A high level framework for general purpose neural networks in Pytorch.__
 
@@ -48,7 +49,7 @@ model = nnt.Sequential(input_shape=1)
 model.add_module('conv1', nnt.Conv2d(model.output_shape, 20, 5, padding='half', activation='relu'))
 model.add_module('conv2', nnt.Conv2d(model.output_shape, 64, 5, padding='half', activation='relu'))
 ```
-which frees you from a lot of memorizations and manual calculations when adding one layer on top of another. 
+which frees you from a lot of memorization and manual calculations when adding one layer on top of another. 
 Theano folks will also find some reminiscence as many functions are highly inspired by Theano.  
 
 # Requirements

diff --git a/tests/function_test.py b/tests/function_test.py
@@ -142,49 +142,51 @@ def test_batch_pairwise_distance(device):
     actual = nnt.utils.batch_pairwise_dist(xyz, xyz, c_code=False)
     testing.assert_allclose(T.diag(actual[0]), T.zeros(actual.shape[1]).to(device))
 
-    if nnt.cuda_ext_available:
-        xyz1 = T.rand(10, 4000, 3).to(device).requires_grad_(True)
-        xyz2 = T.rand(10, 5000, 3).to(device)
-        expected = nnt.utils.batch_pairwise_dist(xyz1, xyz2, c_code=False)
-        actual = nnt.utils.batch_pairwise_dist(xyz1, xyz2, c_code=True)
-        testing.assert_allclose(actual, expected)
+    if dev != 'cuda' or not nnt.cuda_ext_available:
+        pytest.skip('Requires CUDA extension to be installed')
 
-        expected_cost = T.sum(expected)
-        expected_cost.backward()
-        expected_grad = xyz1.grad
-        xyz1.grad.zero_()
+    xyz1 = T.rand(10, 4000, 3).to(device).requires_grad_(True)
+    xyz2 = T.rand(10, 5000, 3).to(device)
+    expected = nnt.utils.batch_pairwise_dist(xyz1, xyz2, c_code=False)
+    actual = nnt.utils.batch_pairwise_dist(xyz1, xyz2, c_code=True)
+    testing.assert_allclose(actual, expected)
 
-        actual_cost = T.sum(actual)
-        actual_cost.backward()
-        actual_grad = xyz1.grad
-        testing.assert_allclose(actual_grad, expected_grad)
+    expected_cost = T.sum(expected)
+    expected_cost.backward()
+    expected_grad = xyz1.grad
+    xyz1.grad.zero_()
 
-        for _ in range(10):
-            t1 = nnt.utils.time_cuda_module(nnt.utils.batch_pairwise_dist, xyz1, xyz2, c_code=False)
-            t2 = nnt.utils.time_cuda_module(nnt.utils.batch_pairwise_dist, xyz1, xyz2, c_code=True)
-            print('pt: %f, cpp: %f' % (t1, t2))
+    actual_cost = T.sum(actual)
+    actual_cost.backward()
+    actual_grad = xyz1.grad
+    testing.assert_allclose(actual_grad, expected_grad)
 
+    for _ in range(10):
+        t1 = nnt.utils.time_cuda_module(nnt.utils.batch_pairwise_dist, xyz1, xyz2, c_code=False)
+        t2 = nnt.utils.time_cuda_module(nnt.utils.batch_pairwise_dist, xyz1, xyz2, c_code=True)
+        print('pt: %f, cpp: %f' % (t1, t2))
 
+
+@pytest.mark.skipif(not nnt.cuda_ext_available, reason='Requires CUDA extension to be installed')
 @pytest.mark.parametrize('device', dev)
 def test_pointcloud_to_voxel(device):
-    if nnt.cuda_ext_available:
-        xyz = T.rand(10, 4000, 3).to(device).requires_grad_(True)
-        pc = xyz * 2. - 1.
-        expected = nnt.utils.pc2vox_fast(pc, c_code=False)
-        actual = nnt.utils.pc2vox_fast(pc, c_code=True)
-        testing.assert_allclose(actual, expected)
-
-        expected_cost = T.sum(expected)
-        expected_cost.backward(retain_graph=True)
-        expected_grad = xyz.grad
-        xyz.grad.zero_()
-
-        actual_cost = T.sum(actual)
-        actual_cost.backward()
-        actual_grad = xyz.grad
-        testing.assert_allclose(actual_grad, expected_grad)
-
-        for _ in range(10):
-            t1 = nnt.utils.time_cuda_module(nnt.utils.pc2vox_fast, pc, c_code=False)
-            t2 = nnt.utils.time_cuda_module(nnt.utils.pc2vox_fast, pc)
-            print('pt: %f, cpp: %f' % (t1, t2))
+    xyz = T.rand(10, 4000, 3).to(device).requires_grad_(True)
+    pc = xyz * 2. - 1.
+    expected = nnt.utils.pc2vox_fast(pc, c_code=False)
+    actual = nnt.utils.pc2vox_fast(pc, c_code=True)
+    testing.assert_allclose(actual, expected)
+
+    expected_cost = T.sum(expected)
+    expected_cost.backward(retain_graph=True)
+    expected_grad = xyz.grad
+    xyz.grad.zero_()
+
+    actual_cost = T.sum(actual)
+    actual_cost.backward()
+    actual_grad = xyz.grad
+    testing.assert_allclose(actual_grad, expected_grad)
+
+    for _ in range(10):
+        t1 = nnt.utils.time_cuda_module(nnt.utils.pc2vox_fast, pc, c_code=False)
+        t2 = nnt.utils.time_cuda_module(nnt.utils.pc2vox_fast, pc)
+        print('pt: %f, cpp: %f' % (t1, t2))
diff --git a/tests/misc_test.py b/tests/misc_test.py
@@ -64,41 +64,29 @@ def test_monitor(device):
     loaded = mon.load('foo.pkl')
     testing.assert_allclose(a, loaded)
 
-    mon.epoch = 0
-    for epoch in mon.iter_epoch(range(n_epochs)):
-        for it in mon.iter_batch(range(n_iters)):
-            mon.dump('foo.pkl', a + it, keep=3)
-            mon.plot('parabol1', (it + epoch) ** 2)
-            mon.hist('histogram1', a + (it * epoch))
-            mon.imwrite('image', a[None, None])
-            mon.plot_matrix('random', a)
-
-    loaded = mon.load('foo.pkl', version=48)
-    testing.assert_allclose(a + 8., loaded)
-
     mon.dump('foo.txt', nnt.utils.to_numpy(a), 'txt')
     loaded = T.from_numpy(mon.load('foo.txt', 'txt', dtype='float32')).to(device)
     testing.assert_allclose(a, loaded)
 
-    mon.epoch = 0
-    for epoch in mon.iter_epoch(range(n_epochs)):
-        for it in mon.iter_batch(range(n_iters)):
-            mon.plot('parabol2', (it + epoch) ** 2)
-            mon.hist('histogram2', a + (it * epoch))
-            mon.dump('foo.txt', nnt.utils.to_numpy(a + it), 'txt', keep=3)
-    loaded = T.from_numpy(mon.load('foo.txt', 'txt', version=48, dtype='float32')).to(device)
-    testing.assert_allclose(a + 8., loaded)
-
     mon.dump('foo.pt', {'a': a}, 'torch')
     loaded = mon.load('foo.pt', 'torch')['a']
     testing.assert_allclose(a, loaded)
 
     mon.epoch = 0
     for epoch in mon.iter_epoch(range(n_epochs)):
         for it in mon.iter_batch(range(n_iters)):
-            mon.plot('parabol3', (it + epoch) ** 2)
-            mon.hist('histogram3', a + (it * epoch))
+            mon.plot('parabol', (it + epoch) ** 2)
+            mon.hist('histogram', a + (it * epoch))
+            mon.imwrite('image', a[None, None])
+            mon.plot_matrix('random', a)
+            mon.dump('foo.pkl', a + it, keep=3)
+            mon.dump('foo.txt', nnt.utils.to_numpy(a + it), 'txt', keep=3)
             mon.dump('foo.pt', {'a': a + it}, 'torch', keep=4)
+
+    loaded = mon.load('foo.pkl', version=48)
+    testing.assert_allclose(a + 8., loaded)
+    loaded = T.from_numpy(mon.load('foo.txt', 'txt', version=48, dtype='float32')).to(device)
+    testing.assert_allclose(a + 8., loaded)
     loaded = mon.load('foo.pt', 'torch', version=49)['a']
     testing.assert_allclose(a + 9, loaded)
 
@@ -143,7 +131,7 @@ def __init__(self, input_shape):
 @pytest.mark.parametrize('pin_memory', (True, False))
 def test_dataloader(device, bs, shuffle, drop_last, pin_memory):
     if device == 'cpu' and pin_memory:
-        pytest.skip('Skip test for pin_memory=True and device=cpu')
+        pytest.skip('Wrong settings for pin_memory=True and device=cpu')
 
     from torch.utils.data import TensorDataset
     data, label = T.arange(10), T.arange(10) + 10