Merge pull request #25 from mlexchange:unit_tests

Unit tests of TiledDataset Class

src/_tests/conftest.py

@@ -33,6 +33,6 @@ def client(context):
     client = from_context(context)
     recons_container = client.create_container("reconstructions")
     recons_container.write_array(np.zeros((2, 3, 3), dtype=np.int8), key="recon1")
-    masks_container = client.create_container("uid0001", metadata={"mask_idx": ["0"]})
-    masks_container.write_array(np.zeros((1, 3, 3), dtype=np.int8), key="mask")
+    masks_container = client.create_container("uid0001", metadata={"mask_idx": ["1"]})
+    masks_container.write_array(np.ones((1, 3, 3), dtype=np.int8), key="mask")
     yield client

src/_tests/example_tunet.yaml

@@ -0,0 +1,41 @@
+# Example for parameters to excecute
+
+# I/O
+io_parameters:
+  data_tiled_uri:
+  data_tiled_api_key:
+  mask_tiled_uri:
+  mask_tiled_api_key:
+  seg_tiled_uri:
+  uid_save:
+  uid_retrieve:
+  models_dir: .
+
+model_parameters:
+  network: "TUNet"
+  num_classes: 3
+  num_epochs: 3
+  optimizer: "Adam"
+  criterion: "CrossEntropyLoss"
+  weights: "[1.0, 2.0, 0.5]"
+  learning_rate: 0.1
+  activation: "ReLU"
+  normalization: "BatchNorm2d"
+  convolution: "Conv2d"
+
+  qlty_window: 64
+  qlty_step: 32
+  qlty_border: 8
+
+  shuffle_train: True
+  batch_size_train: 1
+
+  batch_size_val: 1
+
+  batch_size_inference: 2
+  val_pct: 0.2
+
+  depth: 4
+  base_channels: 8
+  growth_rate: 2
+  hidden_rate: 1

src/_tests/test_tiled_dataset.py

@@ -1,16 +1,52 @@
+import numpy as np
+
 from ..tiled_dataset import TiledDataset
 
 
-def test_tiled_dataset(client):
+def test_with_mask_training(client):
     tiled_dataset = TiledDataset(
-        client["reconstructions"]["recon1"],
+        data_tiled_client=client["reconstructions"]["recon1"],
+        mask_tiled_client=client["uid0001"],
+        is_training=True,
     )
     assert tiled_dataset
+    assert tiled_dataset.mask_idx == [1]
+    assert len(tiled_dataset) == 1
+    assert len(tiled_dataset[0]) == 2
+    # Check data
+    assert tiled_dataset[0][0].shape == (3, 3)
+    assert not np.all(tiled_dataset[0][0])  # should be all 0s
+    # Check mask
+    assert tiled_dataset[0][1].shape == (3, 3)
+    assert np.all(tiled_dataset[0][1])  # should be all 1s
+
+
+def test_with_mask_inference(client):
+    tiled_dataset = TiledDataset(
+        data_tiled_client=client["reconstructions"]["recon1"],
+        mask_tiled_client=client["uid0001"],
+        is_training=False,
+    )
+    assert tiled_dataset
+    assert tiled_dataset.mask_idx == [1]
+    assert len(tiled_dataset) == 1
+    # Check data
     assert tiled_dataset[0].shape == (3, 3)
+    assert not np.all(tiled_dataset[0])  # should be all 0s
 
 
-def test_tiled_dataset_with_masks(client):
+def test_no_mask_inference(client):
     tiled_dataset = TiledDataset(
-        client["reconstructions"]["recon1"], mask_tiled_client=client["uid0001"]
+        data_tiled_client=client["reconstructions"]["recon1"],
+        is_training=False,
     )
+    assert tiled_dataset
+    assert len(tiled_dataset) == 2
+    # Check data
     assert tiled_dataset[0].shape == (3, 3)
+    assert not np.all(tiled_dataset[0])  # should be all 0s
+
+
+# TODO: Test qlty cropping within tiled_dataset.
+# Since this part has been moved to the training script and performed outside,
+# this is not on higher priority.

src/_tests/test_training.py

@@ -0,0 +1,21 @@
+# TODO: test general yaml loading, check param type
+
+# TODO: test model params loading, check pydantic class type / format
+
+# TODO: check dir creation? How to handle file system change during pytest?
+
+# TODO: load TiledDataset from fixture client, test already done.
+
+# TODO: test data and mask array dim and shape
+
+# TODO: test train_loader and val_loader from crop_split_load func, check length
+
+# TODO: test build_network. How to deal with lengthy func? test all network options?
+
+# TODO: test weights and criterion?
+
+# TODO: test dvc?
+
+# TODO: test trainer building
+
+# TODO: test 1 epoch, check param saving