[Python] How to read in a model file saved using mlflow in the scoring script in Azure ML...

I am working on a pipeline that saves a model using mlflow.sklearn.save_model in the training script. This model file then needs to be read in the scoring script

model = LinearRegression()
model.fit(X_train, y_train)

mlflow.sklearn.save_model(
sk_model=model,
path=os.path.join(args.model)
)


The command job is as below:

from azure.ai.ml import command, Input, Output

training_component = command(name=...,
description=...,
display_name=...,
environment=...,
inputs = {
'train_data' : Input(type='uri_folder'),
},
outputs = {
'model' : Output(type='uri_folder', mode='rw_mount')
},
code=training_folder,
command = '''python training.py \
--train_data ${{inputs.train_data}} --model ${{outputs.model}}'''
)


Scoring script is as below:

def main():
# Main scoring script

parser = argparse.ArgumentParser()
parser.add_argument('--test_data', help='test data to check for model accuracy', type=str)
parser.add_argument('--model', help = 'Model to be used for prediction', type = str)
parser.add_argument('--metrics', help='Metrics of the model', type=str)

args = parser.parse_args()

test_data = pd.read_csv(os.path.join(args.test_data, os.listdir(args.test_data)[0]))
model = pd.read_pickle(os.path.join(args.model, os.listdir(args.model)[0]))


In the command job, I am specifying input as:

inputs = {
'test_data' : Input(type='uri_folder'),
'model' : Input(type='uri_folder')
},


but the pipeline fails with below error:

Execution failed. User process 'python' exited with status code 1. Please check log file 'user_logs/std_log.txt' for error details. Error: Traceback (most recent call last):
File "scoring.py", line 40, in <module>
main()
File "scoring.py", line 20, in main
model = pd.read_pickle(os.path.join(args.model, os.listdir(args.model)[0]))
File "/azureml-envs/azureml_bc28677dc89f88f1b17ebecb9a380066/lib/python3.8/site-packages/pandas/io/pickle.py", line 217, in read_pickle
return pickle.load(handles.handle) # type: ignore[arg-type]
_pickle.UnpicklingError: invalid load key, 'f'.


Can someone please let me know how to read in a model file saved using mlflow

Continue reading...