Fix CVEs for Alibi images

.github/workflows/alibiexplainer_tests.yml

@@ -24,10 +24,17 @@ jobs:
   python-tests:
 
     runs-on: ubuntu-latest
-    container: seldonio/python-builder:0.6
-
     steps:
       - uses: actions/checkout@v2
+      - name: Set up Python 3.7
+        uses: actions/setup-python@v2
+        with:
+          python-version: 3.7
+      - name: Install Poetry
+        uses: snok/install-poetry@v1
+        with:
+          version: 1.1.15
+          virtualenvs-create: false
       - name: test-python
         run: |
           pip install --upgrade pip setuptools

components/alibi-detect-server/Dockerfile

@@ -23,10 +23,10 @@ RUN microdnf update -y && \
       gcc-c++
 
 # Install Rclone Binary to be present in the image
-RUN wget https://downloads.rclone.org/v1.55.1/rclone-v1.55.1-linux-amd64.zip && \
-    unzip rclone-v1.55.1-linux-amd64.zip && \
-    mv rclone-v1.55.1-linux-amd64/rclone /usr/bin/rclone && \
-    rm -rf rclone-v1.55.1-linux-amd64.zip rclone-v1.55.1-linux-amd64
+RUN wget https://downloads.rclone.org/v1.61.1/rclone-v1.61.1-linux-amd64.zip && \
+    unzip rclone-v1.61.1-linux-amd64.zip && \
+    mv rclone-v1.61.1-linux-amd64/rclone /usr/bin/rclone && \
+    rm -rf rclone-v1.61.1-linux-amd64.zip rclone-v1.61.1-linux-amd64
 
 # Install Python / Conda
 # Note that we need to force Conda to use the system's std-c++ library, as

components/alibi-detect-server/adserver/__main__.py

@@ -2,9 +2,9 @@
 
 import tensorflow as tf
 
-gpus = tf.config.experimental.list_physical_devices('GPU')
+gpus = tf.config.experimental.list_physical_devices("GPU")
 for gpu in gpus:
-  tf.config.experimental.set_memory_growth(gpu, True)
+    tf.config.experimental.set_memory_growth(gpu, True)
 
 tf.keras.backend.clear_session()
 
@@ -18,8 +18,7 @@
 from adserver.server import CEServer
 from adserver.protocols import Protocol
 from adserver.server import DEFAULT_HTTP_PORT
-from alibi_detect.utils.saving import Data
-
+from adserver.base import Data
 
 
 class AlibiDetectMethod(Enum):

components/alibi-detect-server/adserver/ad_model.py

@@ -5,8 +5,8 @@
 from adserver.constants import HEADER_RETURN_INSTANCE_SCORE
 from .numpy_encoder import NumpyEncoder
 from adserver.protocols.util import read_inputs_as_numpy
-from alibi_detect.utils.saving import load_detector, Data
-from adserver.base import CEModel, ModelResponse
+from alibi_detect.utils.saving import load_detector
+from adserver.base import CEModel, ModelResponse, Data
 from adserver.base.storage import download_model
 
 
@@ -76,5 +76,5 @@ def process_event(self, inputs: Union[List, Dict], headers: Dict) -> ModelRespon
 
         ad_preds = self.model.predict(X, return_instance_score=ret_instance_score)
 
-        data =  json.loads(json.dumps(ad_preds, cls=NumpyEncoder))
+        data = json.loads(json.dumps(ad_preds, cls=NumpyEncoder))
         return ModelResponse(data=data, metrics=None)

components/alibi-detect-server/adserver/base/__init__.py

@@ -1,2 +1,2 @@
 from .model import CEModel, ModelResponse
-from .alibi_model import AlibiDetectModel
+from .alibi_model import AlibiDetectModel, Data

components/alibi-detect-server/adserver/base/alibi_model.py

@@ -1,7 +1,10 @@
-from typing import Optional
+from typing import Optional, Union
 from adserver.base.model import CEModel
 from adserver.base.storage import download_model
-from alibi_detect.utils.saving import load_detector, Data
+from alibi_detect.base import ConfigurableDetector, Detector
+from alibi_detect.utils.saving import load_detector
+
+Data = Union[Detector, ConfigurableDetector]
 
 
 class AlibiDetectModel(CEModel):  # pylint:disable=c-extension-no-member

components/alibi-detect-server/adserver/cd_model.py

@@ -4,12 +4,13 @@
 import numpy as np
 import os
 from .numpy_encoder import NumpyEncoder
-from adserver.base import AlibiDetectModel, ModelResponse
-from alibi_detect.utils.saving import load_detector, Data
+from adserver.base import AlibiDetectModel, ModelResponse, Data
+from alibi_detect.utils.saving import load_detector
 from adserver.constants import ENV_DRIFT_TYPE_FEATURE
 
 DRIFT_TYPE_FEATURE = os.environ.get(ENV_DRIFT_TYPE_FEATURE, "").upper() == "TRUE"
 
+
 def _append_drift_metrcs(metrics, drift, name):
     metric_found = drift.get(name)
 
@@ -58,7 +59,9 @@ def __init__(
         self.batch: Optional[np.ndarray] = None
         self.model: Data = model
 
-    def process_event(self, inputs: Union[List, Dict], headers: Dict) -> Optional[ModelResponse]:
+    def process_event(
+        self, inputs: Union[List, Dict], headers: Dict
+    ) -> Optional[ModelResponse]:
         """
         Process the event and return Alibi Detect score
 
@@ -98,7 +101,7 @@ def process_event(self, inputs: Union[List, Dict], headers: Dict) -> Optional[Mo
                 self.drift_batch_size,
             )
             if DRIFT_TYPE_FEATURE:
-                cd_preds = self.model.predict(self.batch, drift_type='feature')
+                cd_preds = self.model.predict(self.batch, drift_type="feature")
             else:
                 cd_preds = self.model.predict(self.batch)
 

components/alibi-detect-server/adserver/od_model.py

@@ -5,8 +5,8 @@
 import numpy as np
 from .numpy_encoder import NumpyEncoder
 from adserver.protocols.util import read_inputs_as_numpy
-from adserver.base import CEModel, ModelResponse
-from alibi_detect.utils.saving import load_detector, Data
+from adserver.base import CEModel, ModelResponse, Data
+from alibi_detect.utils.saving import load_detector
 from adserver.base.storage import download_model
 from adserver.constants import (
     HEADER_RETURN_INSTANCE_SCORE,
@@ -66,7 +66,9 @@ def load(self):
         self.model: Data = load_detector(model_folder)
         self.ready = True
 
-    def process_event(self, inputs: Union[List, Dict], headers: Dict) -> Optional[ModelResponse]:
+    def process_event(
+        self, inputs: Union[List, Dict], headers: Dict
+    ) -> Optional[ModelResponse]:
         """
         Process the event and return Alibi Detect score
 

components/alibi-detect-server/adserver/tests/test_cd_model.py

@@ -60,7 +60,7 @@ def test_basic(self):
         model = DummyCDModel()
         ad_model = AlibiDetectConceptDriftModel(
             "imdb_text_drift",
-            "gs://seldon-models/alibi-detect/cd/ks/imdb-0_6_2",
+            "gs://seldon-models/alibi-detect/cd/ks/imdb-0_10_4",
             drift_batch_size=2,
         )
         req = [