Merge 7510b1e into cd90979

requirements-test.txt

@@ -1,7 +1,7 @@
 --find-links  https://download.pytorch.org/whl/torch_stable.html
-torch==1.7.1+cpu
-torchvision==0.8.2+cpu
-pytorch_lightning==1.1.0
+torch==1.8.0+cpu
+torchvision==0.9.0+cpu
+pytorch_lightning==1.2.1
 scipy==1.5.4
 timm
 pandas

requirements.txt

@@ -1,4 +1,5 @@
-torch==1.7.1
-torchvision==0.8.2
-pytorch_lightning==1.2.0
+torch==1.8.0
+torchvision==0.9.0
+pytorch_lightning==1.2.1
 scipy==1.5.4
+

tests/test_detr.py

@@ -136,7 +136,7 @@ def test_val_step(self):
 
     def test_fit(self):
         for bbone in some_supported_backbones:
-            backbone = detr.create_detr_backbone(bbone, pretrained="coco")
+            backbone = detr.create_detr_backbone(bbone, pretrained=None)
             self.assertTrue(isinstance(backbone, nn.Module))
             detr_model = detr.create_detr(num_classes=3, num_queries=5, backbone=backbone)
             detr_model = detr_model.cpu()
@@ -156,7 +156,7 @@ def test_fit(self):
     @unittest.skipIf(not torch.cuda.is_available(), "CUDA unavailable")
     def test_fit_cuda(self):
         for bbone in some_supported_backbones:
-            backbone = detr.create_detr_backbone(bbone, pretrained="coco")
+            backbone = detr.create_detr_backbone(bbone, pretrained=None)
             self.assertTrue(isinstance(backbone, nn.Module))
             detr_model = detr.create_detr(num_classes=3, num_queries=5, backbone=backbone)
             self.assertTrue(isinstance(detr_model, nn.Module))
@@ -247,7 +247,7 @@ class LightningTester(unittest.TestCase):
     def test_lit_detr(self):
         flag = False
         for bbone in some_supported_backbones:
-            model = detr.LitDETR(num_classes=3, num_queries=5, pretrained="coco", backbone=bbone)
+            model = detr.LitDETR(num_classes=3, num_queries=5, pretrained=None, backbone=bbone)
             trainer = pl.Trainer(fast_dev_run=True, logger=False, checkpoint_callback=False)
             trainer.fit(model, train_loader, val_loader)
         flag = True

tests/test_layers.py

@@ -0,0 +1,18 @@
+import torch
+import unittest
+import torch.nn as nn
+from quickvision import layers
+
+
+class MishTester(unittest.TestCase):
+    def test_mish(self):
+        pass
+
+
+class MLPTester(unittest.TestCase):
+    def test_mlp(self):
+        pass
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_losses.py

@@ -0,0 +1,35 @@
+import torch
+import torch.nn as nn
+import unittest
+from quickvision import losses
+from quickvision.losses import functional as fl
+
+
+class DiceLossTester(unittest.TestCase):
+    def test_dice_loss_functional(self):
+        # Same thing what you do with below.
+        # inputs = torch.tensor([[0.4, 0.2, 0.3], [0.6, 0.2, 0.3]], dtype=torch.float32)
+        # targets = torch.tensor([[0], [1]], dtype=torch.float32)
+        # loss = fl.dice_loss(inputs, targets)
+        # Do a backward
+        # loss.backward()
+        # And now compare this loss with known valueQ
+        # self.assertTrue()
+        pass
+
+    def test_dice_loss(self):
+        # loss_fn = losses.DiceLoss()
+        # inputs = torch.tensor([[0.4, 0.2, 0.3], [0.6, 0.2, 0.3]], dtype=torch.float32)
+        # targets = torch.tensor([[0], [1]], dtype=torch.float32)
+        # loss = loss_fn(inputs, targets)
+        # See what expected loss should be
+        # expected_loss = []
+        # Compare those two with epsilon
+        # loss.backward()
+        # Assert those with epsilon case
+        # self.assertTrue()
+        pass
+
+
+if __name__ == "__main__":
+    unittest.main()