def evaluate_dicom(model_id: int, orthanc_id: str):
"""
takes in a image, path to a dicom and a eval id
from the db and evaluates the dicom using the image
Args:
model_id (int): the id of the database model
dicom_path (str): the path on the disk to the directory containing the DICOMDIR file
eval_id (int): the database id of the study evaluation that has already been saved by caller
"""
try:
print('evaluating dicom')
eval_id = eval_service.add_evals_to_db(orthanc_id, model_id)
print('getting model')
# get the model from the database
model = model_service.get_model(model_id)
print('downloading dicom')
# download the study from orthanc
study_path, _, _, _, _ = orthanc_service.get_study(orthanc_id)
print('evaluating...')
# evaluate study
eval_service.evaluate(model, [study_path], str(uuid.uuid4()), [eval_id])
except Exception as e:
# catch errors and print output
eval_service.fail_dicom_eval(eval_id)
def evaluate_studies(model_id: List[str], batch_size: int):
"""
Gets studies from orthanc and evaluates all of the applicable studies using a given model
Args:
model_id (:obj:`list`): A list of study IDs from orthanc
batch_size (int): the number of images to process at a time
"""
try:
print(f'evaluating studies for {model_id}')
# get all studies
studies = study_service.get_studies_for_model(model_id, batch_size)
# exit if no studies
if len(studies) < 1:
return
# get the appropriate evaluating model
model = model_service.get_model(model_id)
# get ids of study evaluations
eval_ids = eval_service.get_eval_ids(model, studies) + eval_service.get_failed_eval_ids(model)
eval_service.evaluate_studies(studies, model, eval_ids)
except Exception as e:
eval_service.fail_model(model_id)
def run_experiments(batch_size: int):
"""
Monitors db for active experiments and runs them
"""
print('running experiments')
# get experiments
experiments = experiment_service.get_running_experiments()
# run experiments
for experiment in experiments:
messaging_service.send_notification(f'Started experiment {experiment["name"]}',
'experiment_started')
# restart failed evaluations
eval_service.restart_failed_by_exp(experiment['id'])
# get experiment studies
studies = experiment_service.get_experiment_studies(experiment['id'])
# get model
model = model_service.get_model(experiment['modelId'])
batch = studies[:batch_size]
experiment_service.run_experiment(batch, dict(model), dict(experiment))
from services.exception import TrainingInProgressError
from services.exception import ClassifierNotReadyError
# Globals
classifier_ready = False
is_training = False
is_cleaning = False
MODEL_TYPE = 'STOCKV2'
# Try to load the tokenizer & classfier
classifier = None
tokenizer = None
model_id = model_service.find_current_model_by_model_type(MODEL_TYPE)
if model_id is not None:
model = model_service.get_model(model_id)
classifier = model.predictor
tokenizer = model.tokenizer
classifier_ready = True
# endif
# Begin stockv2 functions --------------------------------------------------------------------------------------------------
def predict(texts):
global classifier_ready
if not classifier_ready:
raise ClassifierNotReadyError()