Added new release version

.github/workflows/codecov.yml

@@ -20,7 +20,7 @@ jobs:
         uses: actions/setup-python@v4
         with:
           cache: 'pip'
-          python-version: "3.9"
+          python-version: "3.10"
 
       - name: Install tox-gh
         run: pip install tox-gh

.github/workflows/python-package.yml

@@ -32,6 +32,8 @@ jobs:
           python-version: ${{ matrix.python-version }}
       - name: Install tox-gh
         run: pip install tox-gh
+      - name: Install quantus with test dependencies
+        run: pip install .[tests]
       - name: Setup test environment
         run: tox run --notest
       - name: Test with PyTest

pyproject.toml

@@ -51,18 +51,24 @@ dynamic = ["version"]
 #
 [project.optional-dependencies]
 torch = [
+    # Mac/Windows or unknown platform fallback
     "torch>=1.13.1; sys_platform != 'linux'",
     "torchvision>=0.15.1; sys_platform != 'linux'",
-    "torch>=1.13.1, <2.0.0; sys_platform == 'linux' and python_version <= '3.10'",
-    "torchvision>=0.14.0, <0.15.1; sys_platform == 'linux' and python_version <= '3.10'",
+
+    # Linux + Python 3.8–3.10 → use torch 1.x
+    "torch>=1.13.1, <2.0.0; sys_platform == 'linux' and python_version < '3.11'",
+    "torchvision>=0.14.0, <0.15.1; sys_platform == 'linux' and python_version < '3.11'",
+
+    # Linux + Python 3.11+ → use torch 2.x
     "torch>=2.0.0; sys_platform == 'linux' and python_version >= '3.11'",
     "torchvision>=0.15.1; sys_platform == 'linux' and python_version >= '3.11'",
 ]
 tensorflow = [
     # 2.16 is shipped without keras
     "tensorflow<2.16.0",
     # keras V3 broke everything
-    "keras<3"
+    "keras<3",
+    "numpy<2"
 ]
 captum = [
     "quantus[torch]",
@@ -78,10 +84,10 @@ zennit = [
 ]
 transformers = [
     "quantus[torch]",
-    "transformers>=4.38.2",
+    "transformers>=4.38.2", #"transformers<4.38.0"
 ]
 full = [
-    "quantus[captum,tf-explain,zennit,transformers]"
+    "quantus[captum,tf-explain,zennit,transformers,torch]"
 ]
 tests = [
     "coverage>=7.2.3",
@@ -91,6 +97,7 @@ tests = [
     "pytest-lazy-fixture>=0.6.3",
     "pytest-mock==3.10.0",
     "pytest_xdist",
+    "numpy<2",
     "quantus[full]"
 ]
 

quantus/__init__.py

@@ -5,7 +5,7 @@
 # Quantus project URL: <https://github.com/understandable-machine-intelligence-lab/Quantus>.
 
 # Set the correct version.
-__version__ = "0.5.3"
+__version__ = "0.6.0"
 
 # Expose quantus.evaluate to the user.
 from quantus.evaluation import evaluate

quantus/metrics/faithfulness/selectivity.py

@@ -375,6 +375,6 @@ def evaluate_batch(
             y_pred_perturb = model.predict(x_input)[np.arange(batch_size), y_batch]
 
             results.append(y_pred_perturb)
-        results = np.stack(results, 1, dtype=np.float64)
+        results = np.stack(results, 1).astype(np.float64)
 
         return results

tests/conftest.py

@@ -239,23 +239,36 @@ def load_mnist_model_softmax():
 
 @pytest.fixture(scope="session", autouse=False)
 def load_hf_distilbert_sequence_classifier():
-    """
-    TODO
-    """
-    DISTILBERT_BASE = "distilbert-base-uncased"
-    model = AutoModelForSequenceClassification.from_pretrained(DISTILBERT_BASE, cache_dir="/tmp/")
-    return model
+    try:
+        import torch
+    except ImportError:
+        pytest.skip("Skipping because torch is not available.")
+
+    try:
+        from transformers import AutoModelForSequenceClassification
+        DISTILBERT_BASE = "distilbert-base-uncased"
+        return AutoModelForSequenceClassification.from_pretrained(DISTILBERT_BASE, cache_dir="/tmp/")
+    except Exception as e:
+        pytest.skip(f"Skipping because model loading failed: {e}")
+
 
 
 @pytest.fixture(scope="session", autouse=False)
 def dummy_hf_tokenizer():
-    """
-    TODO
-    """
-    DISTILBERT_BASE = "distilbert-base-uncased"
-    REFERENCE_TEXT = "The quick brown fox jumps over the lazy dog"
-    tokenizer = AutoTokenizer.from_pretrained(DISTILBERT_BASE, cache_dir="/tmp/", clean_up_tokenization_spaces=True)
-    return tokenizer(REFERENCE_TEXT, return_tensors="pt")
+    try:
+        import torch
+    except ImportError:
+        pytest.skip("Skipping because torch is not available.")
+
+    try:
+        from transformers import AutoTokenizer
+        DISTILBERT_BASE = "distilbert-base-uncased"
+        REFERENCE_TEXT = "The quick brown fox jumps over the lazy dog"
+        tokenizer = AutoTokenizer.from_pretrained(DISTILBERT_BASE, cache_dir="/tmp/", clean_up_tokenization_spaces=True)
+        return tokenizer(REFERENCE_TEXT, return_tensors="pt")
+    except Exception as e:
+        pytest.skip(f"Skipping because tokenizer loading failed: {e}")
+
 
 
 @pytest.fixture(scope="session", autouse=True)

tests/functions/test_pytorch_model.py

@@ -11,6 +11,12 @@
 from scipy.special import softmax
 from quantus.helpers.model.pytorch_model import PyTorchModel
 
+def torch_available():
+    try:
+        import torch
+        return True
+    except ImportError:
+        return False
 
 @pytest.fixture
 def mock_input_torch_array():

tox.ini

@@ -12,12 +12,17 @@ deps =
     .[tests]
 pass_env =
     TF_XLA_FLAGS
+commands_pre =
+    python -c "import sys, platform; print(f'Python: {sys.version_info.major}.{sys.version_info.minor}, Platform: {platform.system()}')"
+    python -c "import sys; import subprocess; v=sys.version_info; linux=sys.platform=='linux'; needs=torch_install=(linux and (v.major==3 and v.minor in (9,10))); subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'torch<2.0.0']) if needs else None"
 commands =
     pytest -s -v {posargs}
 
+
 [testenv:coverage]
 description = Run the tests with coverage
-base_python = py310
+base_python =
+
 deps =
     {[testenv]deps}
     coverage
@@ -27,14 +32,14 @@ commands =
 
 [testenv:build]
 description = Build environment
-base_python = py310
 deps =
-    .
     build
     twine
 commands =
-    python3 -m build .
-    python3 -m twine check ./dist/* --strict
+    python -c "import shutil; shutil.rmtree('dist', ignore_errors=True)"
+    python -m build .
+    python -m twine check ./dist/* --strict
+
 
 [testenv:lint]
 description = Check the code style