Fix tests

Author: sachanub

buildspec.yml

@@ -30,23 +30,23 @@ phases:
       - PY_COLORS=0
 
       # run linters
-      # - tox -e flake8,twine
+      - tox -e flake8,twine
 
       # run unit tests
-      # - tox -e py38,py39,py310 test/unit
+      - tox -e py38,py39,py310 test/unit
 
       # define EIA tag
       # - DLC_EIA_TAG="$EIA_FRAMEWORK_VERSION-dlc-eia-$BUILD_ID"
 
       # run local CPU integration tests (build and push the image to ECR repo)
-      #- |
-      #  for FRAMEWORK_VERSION in $FRAMEWORK_VERSIONS; 
-      #    do
-      #      DLC_CPU_TAG="$FRAMEWORK_VERSION-dlc-cpu-$BUILD_ID";
-      #      test_cmd="IGNORE_COVERAGE=- tox -e py38 -- test/integration/local -vv -rA -s --build-image --push-image --dockerfile-type dlc.cpu --region $AWS_DEFAULT_REGION --docker-base-name $ECR_REPO --aws-id $ACCOUNT --framework-version $FRAMEWORK_VERSION --processor cpu --tag $DLC_CPU_TAG";
-      #      execute-command-if-has-matching-changes "$test_cmd" "test/" "src/*.py" "setup.py" "setup.cfg";
-      #      docker system prune --all --force;
-      #    done
+      - |
+        for FRAMEWORK_VERSION in $FRAMEWORK_VERSIONS; 
+          do
+            DLC_CPU_TAG="$FRAMEWORK_VERSION-dlc-cpu-$BUILD_ID";
+            test_cmd="IGNORE_COVERAGE=- tox -e py38 -- test/integration/local -vv -rA -s --build-image --push-image --dockerfile-type dlc.cpu --region $AWS_DEFAULT_REGION --docker-base-name $ECR_REPO --aws-id $ACCOUNT --framework-version $FRAMEWORK_VERSION --processor cpu --tag $DLC_CPU_TAG";
+            execute-command-if-has-matching-changes "$test_cmd" "test/" "src/*.py" "setup.py" "setup.cfg";
+            docker system prune --all --force;
+          done
 
       # launch remote GPU instance with Deep Learning AMI GPU PyTorch 1.9 (Ubuntu 20.04)
       - prefix='ml.'
@@ -110,4 +110,4 @@ phases:
             aws ecr batch-delete-image --repository-name $ECR_REPO --region $AWS_DEFAULT_REGION --image-ids imageTag=$DLC_GPU_TAG;
           done
       
-      # - aws ecr batch-delete-image --repository-name $ECR_REPO --region $AWS_DEFAULT_REGION --image-ids imageTag=$DLC_EIA_TAG
+      # - aws ecr batch-delete-image --repository-name $ECR_REPO --region $AWS_DEFAULT_REGION --image-ids imageTag=$DLC_EIA_TAG

test/resources/mnist/model_cpu/1d/code/mnist_1d.py

@@ -11,7 +11,6 @@
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
 from __future__ import absolute_import
-
 import os
 import torch
 import torch.nn as nn
@@ -45,4 +44,4 @@ def model_fn(model_dir):
     model = torch.nn.DataParallel(Net())
     with open(os.path.join(model_dir, 'torch_model.pth'), 'rb') as f:
         model.load_state_dict(torch.load(f))
-    return model
+    return model

test/resources/mnist/model_cpu/code/call_model_fn_once.py

@@ -14,6 +14,7 @@
 
 import os
 
+
 def model_fn(model_dir):
     lock_file = os.path.join(model_dir, 'model_fn.lock.{}'.format(os.getpid()))
     if os.path.exists(lock_file):
@@ -33,4 +34,4 @@ def predict_fn(data, model):
 
 
 def output_fn(prediction, accept):
-    return prediction
+    return prediction

test/resources/mnist/model_gpu/code/mnist.py

@@ -63,4 +63,4 @@ def model_fn(model_dir):
     # Move the model to the GPU
     device = torch.device("cuda")
     model = model.to(device)
-    return model
+    return model 

test/resources/model_gpu_context/code/inference.py

@@ -12,12 +12,14 @@
 # language governing permissions and limitations under the License.
 from __future__ import absolute_import
 
+import os
 import torch
 
 def model_fn(model_dir, context):
-    device = torch.device("cuda:" + str(context.system_properties.get("gpu_id")))
-    file_path = "device_info.txt"
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+    file_path = os.path.join(script_dir, "device_info.txt")
 
+    device = torch.device("cuda:" + str(context.system_properties.get("gpu_id")))
     device_str = str(device)[-1]
     with open(file_path, "a") as file:
         file.write(device_str + "\n")