import mlflow
import mlflow.pytorch
from mlflow.tracking import MlflowClient
import numpy as np
import torch
import json

#mlflow.set_tracking_uri("http://127.0.0.1:5000")
mlflow.set_tracking_uri("http://172.17.0.1:5000")
client = MlflowClient()
version = 0
model_name = "s426206"
for mv in client.search_model_versions("name='s426206'"):
    if int(mv.version) > version:
        version = int(mv.version)

model = mlflow.pytorch.load_model(
    model_uri=f"models:/{model_name}/{version}"
)

with open('my_model/input_example.json') as json_file:
    data = json.load(json_file)
#print(np.array(data['inputs']))
print(model(torch.tensor(np.array(data['inputs'])).float()))