switched to CPU for torch tests

.gitlab-ci.yml

@@ -2,16 +2,11 @@ image: python:3.7
 
 tests:
   before_script:
-    - date
     - curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python
     - source $HOME/.poetry/env
     - poetry config virtualenvs.create false
-    - date
     - poetry install
-    - date
     - python -m nltk.downloader punkt
-    - date
     - python -c 'import stanza; stanza.download("en")'
-    - date
   script:
     - pytest tests

tests/test_torch.py

@@ -10,7 +10,9 @@ from tuhlbox.torch_classifier import TorchClassifier
 from tuhlbox.torch_cnn import CharCNN
 from tuhlbox.torch_lstm import RNNClassifier
 
-x, y = fetch_20newsgroups(return_X_y=True)
+x, y = fetch_20newsgroups(
+    return_X_y=True, categories=["alt.atheism", "talk.religion.misc"]
+)
 
 x_train, x_test, y_train, y_test = train_test_split(x, y)
 
@@ -25,7 +27,7 @@ def test_cnn() -> None:
         Padder2d(pad_value=VOCAB_SIZE, max_len=MAX_SEQ_LEN, dtype=int),
         TorchClassifier(
             module=CharCNN,
-            device="cuda",
+            device="cpu",  # the gitlab CI does not have cuda
             batch_size=54,
             max_epochs=5,
             learn_rate=0.01,
@@ -39,8 +41,7 @@ def test_cnn() -> None:
     )
 
     pipe.fit(x_train, y_train)
-    predictions = pipe.predict(x_test)
-    print(accuracy_score(predictions, y_test))
+    pipe.predict(x_test)
 
 
 def test_lstm() -> None:
@@ -49,8 +50,8 @@ def test_lstm() -> None:
         Padder2d(pad_value=VOCAB_SIZE, max_len=MAX_SEQ_LEN, dtype=int),
         TorchClassifier(
             module=RNNClassifier,
-            device="cuda",
-            batch_size=54,
+            device="cpu",  # the gitlab CI does not have cuda
+            batch_size=4,
             max_epochs=5,
             learn_rate=0.01,
             optimizer=torch.optim.Adam,
@@ -58,5 +59,4 @@ def test_lstm() -> None:
     )
 
     pipe.fit(x_train, y_train)
-    predictions = pipe.predict(x_test)
-    print(accuracy_score(predictions, y_test))
+    pipe.predict(x_test)