Merge pull request #730 from NatLibFi/update-dependencies-v1.0-keras-…

…save-format

Switch to Keras v3 save format for nn_ensemble

annif/backend/nn_ensemble.py

@@ -8,13 +8,14 @@
 from typing import TYPE_CHECKING, Any
 
 import joblib
+import keras.backend as K
 import lmdb
 import numpy as np
-import tensorflow.keras.backend as K
+from keras.layers import Add, Dense, Dropout, Flatten, Input, Layer
+from keras.models import Model
+from keras.saving import load_model
+from keras.utils import Sequence
 from scipy.sparse import csc_matrix, csr_matrix
-from tensorflow.keras.layers import Add, Dense, Dropout, Flatten, Input, Layer
-from tensorflow.keras.models import Model, load_model
-from tensorflow.keras.utils import Sequence
 
 import annif.corpus
 import annif.parallel
@@ -97,7 +98,7 @@ class NNEnsembleBackend(backend.AnnifLearningBackend, ensemble.BaseEnsembleBacke
 
     name = "nn_ensemble"
 
-    MODEL_FILE = "nn-model.h5"
+    MODEL_FILE = "nn-model.keras"
     LMDB_FILE = "nn-train.mdb"
 
     DEFAULT_PARAMETERS = {

tests/test_backend_nn_ensemble.py

@@ -105,11 +105,11 @@ def test_nn_ensemble_train_and_learn(registry, tmpdir):
     assert nn_ensemble._model.optimizer.learning_rate.value() == 0.001
 
     datadir = py.path.local(project.datadir)
-    assert datadir.join("nn-model.h5").exists()
-    assert datadir.join("nn-model.h5").size() > 0
+    assert datadir.join("nn-model.keras").exists()
+    assert datadir.join("nn-model.keras").size() > 0
 
     # test online learning
-    modelfile = datadir.join("nn-model.h5")
+    modelfile = datadir.join("nn-model.keras")
 
     old_size = modelfile.size()
     old_mtime = modelfile.mtime()
@@ -129,7 +129,7 @@ def test_nn_ensemble_train_cached(registry):
     datadir = py.path.local(project.datadir)
     assert datadir.join("nn-train.mdb").exists()
 
-    datadir.join("nn-model.h5").remove()
+    datadir.join("nn-model.keras").remove()
 
     nn_ensemble_type = annif.backend.get_backend("nn_ensemble")
     nn_ensemble = nn_ensemble_type(
@@ -140,8 +140,8 @@ def test_nn_ensemble_train_cached(registry):
 
     nn_ensemble.train("cached")
 
-    assert datadir.join("nn-model.h5").exists()
-    assert datadir.join("nn-model.h5").size() > 0
+    assert datadir.join("nn-model.keras").exists()
+    assert datadir.join("nn-model.keras").size() > 0
 
 
 def test_nn_ensemble_train_and_learn_params(registry, tmpdir, capfd):