def test_sklearn_mnist_classification(self):
test_pipeline = TestPipeline(is_integration_test=False)
input_file = 'gs://apache-beam-ml/testing/inputs/it_mnist_data.csv'
output_file_dir = 'gs://temp-storage-for-end-to-end-tests'
output_file = '/'.join(
[output_file_dir, str(uuid.uuid4()), 'result.txt'])
model_path = 'gs://apache-beam-ml/models/mnist_model_svm.pickle'
extra_opts = {
'input': input_file,
'output': output_file,
'model_path': model_path,
}
sklearn_mnist_classification.run(
test_pipeline.get_full_options_as_args(**extra_opts),
save_main_session=False)
self.assertEqual(FileSystems().exists(output_file), True)
expected_output_filepath = 'gs://apache-beam-ml/testing/expected_outputs/test_sklearn_mnist_classification_actuals.txt' # pylint: disable=line-too-long
expected_outputs = process_outputs(expected_output_filepath)
predicted_outputs = process_outputs(output_file)
self.assertEqual(len(expected_outputs), len(predicted_outputs))
predictions_dict = {}
for i in range(len(predicted_outputs)):
true_label, prediction = predicted_outputs[i].split(',')
predictions_dict[true_label] = prediction
for i in range(len(expected_outputs)):
true_label, expected_prediction = expected_outputs[i].split(',')
self.assertEqual(predictions_dict[true_label], expected_prediction)
def read_image(image_file_name: str,
path_to_dir: Optional[str] = None) -> Tuple[str, Image.Image]:
if path_to_dir is not None:
image_file_name = os.path.join(path_to_dir, image_file_name)
with FileSystems().open(image_file_name, 'r') as file:
data = Image.open(io.BytesIO(file.read())).convert('RGB')
return image_file_name, data
def test_torch_run_inference_bert_for_masked_lm(self):
test_pipeline = TestPipeline(is_integration_test=True)
# Path to text file containing some sentences
file_of_sentences = 'gs://apache-beam-ml/datasets/custom/sentences.txt' # disable: line-too-long
output_file_dir = 'gs://apache-beam-ml/testing/predictions'
output_file = '/'.join(
[output_file_dir, str(uuid.uuid4()), 'result.txt'])
model_state_dict_path = 'gs://apache-beam-ml/models/huggingface.BertForMaskedLM.bert-base-uncased.pth'
extra_opts = {
'input': file_of_sentences,
'output': output_file,
'model_state_dict_path': model_state_dict_path,
}
pytorch_language_modeling.run(
test_pipeline.get_full_options_as_args(**extra_opts),
save_main_session=False)
self.assertEqual(FileSystems().exists(output_file), True)
predictions = process_outputs(filepath=output_file)
actuals_file = 'gs://apache-beam-ml/testing/expected_outputs/test_torch_run_inference_bert_for_masked_lm_actuals.txt'
actuals = process_outputs(filepath=actuals_file)
predictions_dict = {}
for prediction in predictions:
text, predicted_text = prediction.split(';')
predictions_dict[text] = predicted_text
for actual in actuals:
text, actual_predicted_text = actual.split(';')
predicted_predicted_text = predictions_dict[text]
self.assertEqual(actual_predicted_text, predicted_predicted_text)
def test_torch_run_inference_coco_maskrcnn_resnet50_fpn(self):
test_pipeline = TestPipeline(is_integration_test=True)
# text files containing absolute path to the coco validation data on GCS
file_of_image_names = 'gs://apache-beam-ml/testing/inputs/it_coco_validation_inputs.txt' # disable: line-too-long
output_file_dir = 'gs://apache-beam-ml/testing/predictions'
output_file = '/'.join(
[output_file_dir, str(uuid.uuid4()), 'result.txt'])
model_state_dict_path = 'gs://apache-beam-ml/models/torchvision.models.detection.maskrcnn_resnet50_fpn.pth'
images_dir = 'gs://apache-beam-ml/datasets/coco/raw-data/val2017'
extra_opts = {
'input': file_of_image_names,
'output': output_file,
'model_state_dict_path': model_state_dict_path,
'images_dir': images_dir,
}
pytorch_image_segmentation.run(
test_pipeline.get_full_options_as_args(**extra_opts),
save_main_session=False)
self.assertEqual(FileSystems().exists(output_file), True)
predictions = process_outputs(filepath=output_file)
actuals_file = 'gs://apache-beam-ml/testing/expected_outputs/test_torch_run_inference_coco_maskrcnn_resnet50_fpn_actuals.txt'
actuals = process_outputs(filepath=actuals_file)
predictions_dict = {}
for prediction in predictions:
filename, prediction_labels = prediction.split(';')
predictions_dict[filename] = prediction_labels
for actual in actuals:
filename, actual_labels = actual.split(';')
prediction_labels = predictions_dict[filename]
self.assertEqual(actual_labels, prediction_labels)
def _open_dataset(self, path: str) -> xarray.Dataset:
"""Open as an XArray Dataset, sometimes with local caching."""
if self.local_copy:
with tempfile.TemporaryDirectory() as tmpdir:
local_file = fsspec.open_local(
f"simplecache::{path}",
simplecache={'cache_storage': tmpdir}
)
yield xarray.open_dataset(local_file, **self.xarray_open_kwargs)
else:
with FileSystems().open(path) as file:
yield xarray.open_dataset(file, **self.xarray_open_kwargs)
def test_torch_run_inference_imagenet_mobilenetv2(self):
test_pipeline = TestPipeline(is_integration_test=True)
# text files containing absolute path to the imagenet validation data on GCS
file_of_image_names = 'gs://apache-beam-ml/testing/inputs/it_mobilenetv2_imagenet_validation_inputs.txt' # disable: line-too-long
output_file_dir = 'gs://apache-beam-ml/testing/predictions'
output_file = '/'.join(
[output_file_dir, str(uuid.uuid4()), 'result.txt'])
model_state_dict_path = 'gs://apache-beam-ml/models/imagenet_classification_mobilenet_v2.pt'
extra_opts = {
'input': file_of_image_names,
'output': output_file,
'model_state_dict_path': model_state_dict_path,
}
pytorch_image_classification.run(
test_pipeline.get_full_options_as_args(**extra_opts),
save_main_session=False)
self.assertEqual(FileSystems().exists(output_file), True)
predictions = process_outputs(filepath=output_file)
for prediction in predictions:
filename, prediction = prediction.split(',')
self.assertEqual(_EXPECTED_OUTPUTS[filename], prediction)
def process_outputs(filepath):
with FileSystems().open(filepath) as f:
lines = f.readlines()
lines = [l.decode('utf-8').strip('\n') for l in lines]
return lines
