use a jit-disable plugin

azure-pipelines.yml

@@ -50,7 +50,7 @@ jobs:
     displayName: 'Test LKPY'
 
   - script: |
-      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval --no-slow --verbose
+      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval
     displayName: 'Test Coverage'
 
   - script: |
@@ -112,7 +112,7 @@ jobs:
     displayName: 'pytest'
   
   - script: |
-      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval --no-slow --verbose
+      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval
     displayName: 'Test Coverage'
 
   - script: |
@@ -156,7 +156,7 @@ jobs:
     displayName: 'pytest'
   
   - script: |
-      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval --no-slow --verbose
+      env NUMBA_DISABLE_JIT=1 invoke test --cover --no-eval
     displayName: 'Test Coverage'
 
   - script: |

tests/lk_test_utils.py

@@ -89,6 +89,10 @@ def ml_sample():
 ml100k = _ML100K()
 
 
+wantjit = pytest.mark.skipif('NUMBA_DISABLE_JIT' in os.environ,
+                             reason='JIT required')
+
+
 @contextmanager
 def envvars(**vars):
     save = {}

tests/test_als_explicit.py

@@ -85,7 +85,7 @@ def test_als_predict_bad_user():
     assert np.isnan(preds.loc[3])
 
 
-@mark.slow
+@lktu.wantjit
 def test_als_train_large():
     algo = als.BiasedMF(20, iterations=10)
     ratings = lktu.ml_pandas.renamed.ratings

tests/test_als_implicit.py

@@ -58,7 +58,7 @@ def test_als_predict_bad_user():
     assert np.isnan(preds.loc[3])
 
 
-@mark.slow
+@lktu.wantjit
 def test_als_train_large():
     algo = als.ImplicitMF(20, iterations=20)
     ratings = lktu.ml_pandas.renamed.ratings
@@ -88,7 +88,7 @@ def test_als_save_load(tmp_path):
     assert np.all(restored.user_index_ == algo.user_index_)
 
 
-@mark.slow
+@lktu.wantjit
 def test_als_train_large_noratings():
     algo = als.ImplicitMF(20, iterations=20)
     ratings = lktu.ml_pandas.renamed.ratings

tests/test_batch_sweep.py

@@ -99,7 +99,7 @@ def test_sweep_persist(tmp_path):
     sweep.add_datasets(lambda: xf.partition_users(ratings, 5, xf.SampleN(5)), name='ml-small')
     sweep.persist_data()
 
-    for i in range(1,6):
+    for i in range(1, 6):
         assert (work / 'ds{}-train.parquet'.format(i)).exists()
         assert (work / 'ds{}-test.parquet'.format(i)).exists()
 

tests/test_funksvd.py

@@ -123,6 +123,7 @@ def test_fsvd_predict_bad_user():
     assert np.isnan(preds.loc[3])
 
 
+@lktu.wantjit
 @mark.slow
 def test_fsvd_save_load(tmp_path):
     tmp_path = lktu.norm_path(tmp_path)
@@ -151,6 +152,7 @@ def test_fsvd_save_load(tmp_path):
     assert np.all(algo.user_index_ == original.user_index_)
 
 
+@lktu.wantjit
 @mark.slow
 def test_fsvd_known_preds():
     algo = svd.FunkSVD(15, iterations=125, lrate=0.001)
@@ -178,6 +180,7 @@ def test_fsvd_known_preds():
         raise e
 
 
+@lktu.wantjit
 @mark.slow
 @mark.eval
 @mark.skipif(not lktu.ml100k.available, reason='ML100K data not present')

tests/test_knn_item_item.py

@@ -114,7 +114,7 @@ def test_ii_simple_implicit_predict():
     assert res.loc[6] > 0
 
 
-@mark.slow
+@lktu.wantjit
 @mark.skip("redundant with large_models")
 def test_ii_train_big():
     "Simple tests for bounded models"
@@ -132,7 +132,7 @@ def test_ii_train_big():
     assert means[algo.item_index_].values == approx(algo.item_means_)
 
 
-@mark.slow
+@lktu.wantjit
 @mark.skip("redundant with large_models")
 def test_ii_train_big_unbounded():
     "Simple tests for unbounded models"
@@ -150,7 +150,7 @@ def test_ii_train_big_unbounded():
     assert means[algo.item_index_].values == approx(algo.item_means_)
 
 
-@mark.slow
+@lktu.wantjit
 @mark.skipif(not lktu.ml100k.available, reason='ML100K data not present')
 def test_ii_train_ml100k(tmp_path):
     "Test an unbounded model on ML-100K"
@@ -197,7 +197,7 @@ def test_ii_train_ml100k(tmp_path):
         assert all(r_mat.values[sp:ep] == o_mat.values[sp:ep])
 
 
-@mark.slow
+@lktu.wantjit
 def test_ii_large_models():
     "Several tests of large trained I-I models"
     _log.info('training limited model')
@@ -291,6 +291,7 @@ def test_ii_large_models():
             assert all(ubs_except_min.index.isin(b_nbrs.index))
 
 
+@lktu.wantjit
 def test_ii_save_load(tmp_path):
     "Save and load a model"
     tmp_path = lktu.norm_path(tmp_path)
@@ -404,6 +405,7 @@ def test_ii_implicit_save_load(tmp_path):
         assert all(np.diff(row.data) < 1.0e-6)
 
 
+@lktu.wantjit
 @mark.slow
 def test_ii_implicit():
     algo = knn.ItemItem(20, save_nbrs=100, center=False, aggregate='sum')
@@ -448,7 +450,7 @@ def eval(train, test):
     assert user_rmse.mean() == approx(0.90, abs=0.05)
 
 
-@mark.slow
+@lktu.wantjit
 def test_ii_known_preds():
     from lenskit import batch
 
@@ -480,6 +482,7 @@ def test_ii_known_preds():
         raise e
 
 
+@lktu.wantjit
 @mark.slow
 @mark.eval
 @mark.parametrize('ncpus', [1, 2])