fix path in jobs and specs (#143)

* fix path in jobs and specs
* add PYTHONPATH hack to both runnable files
* ignore E402

benchmark/src/components/pytorch_benchmark/image_classifier.py

@@ -12,6 +12,7 @@ Using your editor, search for those strings to get an idea of how to implement:
 - MLFLOW : how to implement mlflow reporting of metrics and artifacts
 - PROFILER : how to implement pytorch profiler
 """
+import os
 import sys
 import time
 import logging
@@ -24,16 +25,22 @@ import mlflow
 # the long list of torch imports
 import torch
 
+# fix to AzureML PYTHONPATH
+ROOT_FOLDER_PATH = os.path.join(os.path.dirname(__file__), "..")
+if ROOT_FOLDER_PATH not in sys.path:
+    print(f"Adding root folder to PYTHONPATH: {ROOT_FOLDER_PATH}")
+    sys.path.append(ROOT_FOLDER_PATH)
+
 # internal imports
 ## non-specific helper code
-from common.profiling import LogTimeBlock, LogDiskIOBlock
+from common.profiling import LogTimeBlock, LogDiskIOBlock  # noqa : E402
 
 ## pytorch generic helping code
-from pytorch_benchmark.helper.training import PyTorchDistributedModelTrainingSequence
+from pytorch_benchmark.helper.training import PyTorchDistributedModelTrainingSequence  # noqa : E402
 
 ## classification specific code
-from pytorch_benchmark.classification.model import get_model_metadata, load_model
-from pytorch_benchmark.classification.io import build_image_datasets
+from pytorch_benchmark.classification.model import get_model_metadata, load_model  # noqa : E402
+from pytorch_benchmark.classification.io import build_image_datasets  # noqa : E402
 
 SCRIPT_START_TIME = time.time()  # just to measure time to start
 

benchmark/src/components/tensorflow_benchmark/image_segmentation.py

@@ -11,6 +11,8 @@ Using your editor, search for those strings to get an idea of how to implement:
 - DISTRIBUTED : how to implement distributed tensorflow
 - MLFLOW : how to implement mlflow reporting of metrics and artifacts
 """
+import os
+import sys
 import time
 import logging
 import argparse
@@ -22,16 +24,22 @@ import mlflow
 import tensorflow as tf
 from tensorflow import keras
 
+# fix to AzureML PYTHONPATH
+ROOT_FOLDER_PATH = os.path.join(os.path.dirname(__file__), "..")
+if ROOT_FOLDER_PATH not in sys.path:
+    print(f"Adding root folder to PYTHONPATH: {ROOT_FOLDER_PATH}")
+    sys.path.append(ROOT_FOLDER_PATH)
+
 # internal imports
 ## non-specific helper code
-from common.profiling import LogTimeBlock, LogDiskIOBlock
+from common.profiling import LogTimeBlock, LogDiskIOBlock  # noqa : E402
 
 ## tensorflow generic helping code
-from tensorflow_benchmark.helper.training import TensorflowDistributedModelTrainingSequence
+from tensorflow_benchmark.helper.training import TensorflowDistributedModelTrainingSequence  # noqa : E402
 
 ## classification specific code
-from tensorflow_benchmark.segmentation.model import load_model
-from tensorflow_benchmark.segmentation.io import ImageAndMaskSequenceDataset
+from tensorflow_benchmark.segmentation.model import load_model  # noqa : E402
+from tensorflow_benchmark.segmentation.io import ImageAndMaskSequenceDataset  # noqa : E402
 
 
 SCRIPT_START_TIME = time.time()  # just to measure time to start

benchmark/src/components/tensorflow_benchmark/image_segmentation_spec.yaml

@@ -104,7 +104,7 @@ code: ../
 environment: azureml:nvidia_tensorflow:22.02-tf2-py3-mod3
 
 command: >-
-  python train.py 
+  python tensorflow_benchmark/image_segmentation.py 
   --train_images ${{inputs.train_images}}
   --train_masks ${{inputs.train_masks}}
   --test_images ${{inputs.test_images}}

benchmark/src/jobs/benchmark/tf_unet_pets.yml

@@ -1,5 +1,5 @@
 $schema: https://azuremlschemas.azureedge.net/latest/commandJob.schema.json
-code: ../../components/tensorflow_image_segmentation/
+code: ../../components/
 
 display_name: "tf_unet"
 experiment_name: "tensorflow_unet_pets"
@@ -73,7 +73,7 @@ outputs:
 ###############
 
 command: >-
-    python train.py 
+    python tensorflow_benchmark/image_segmentation.py 
     --train_images ${{inputs.train_images}}
     --train_masks ${{inputs.train_masks}}
     --test_images ${{inputs.test_images}}

benchmark/src/jobs/benchmark/tf_unet_pets_sweep_datafetching.yml

@@ -14,12 +14,12 @@ compute: azureml:gpu-cluster
 
 
 trial:
-    code: ../../components/tensorflow_image_segmentation/
+    code: ../../components/
 
     ### COMMAND ###
 
     command: >-
-        python train.py 
+        python tensorflow_benchmark/image_segmentation.py 
         --train_images ${{inputs.train_images}}
         --train_masks ${{inputs.train_masks}}
         --test_images ${{inputs.test_images}}