main.py

from typing import List,Dict
from pydantic import BaseModel
import numpy as np 
from fastapi import FastAPI
import torch
from model import BinaryClassificationWithLogits
import __main__

model_path="classification_gaussian_binary_model_0v.pt"
model=BinaryClassificationWithLogits(in_features=4,
                                    out_features=1,
                                    hidden_features=10)

model = torch.jit.load(model_path,map_location="cpu")


class ClassificationFeatures(BaseModel):
  feature_1:float
  feature_2:float
  feature_3:float
  feature_4:float


# Creando una instacnia de FastAPI
app=FastAPI()
# Definiendo la ruta raiz
@app.get("/")
def home_page():
  return "Welcome the API with pytorch"
# Definiendo ruta para inferencias
@app.post("/predict")
def predict_sample(cls_features:ClassificationFeatures) -> Dict:
  
  input_data=np.array([[
        cls_features.feature_1,
        cls_features.feature_2,
        cls_features.feature_3,
        cls_features.feature_4,
    ]])
  X=torch.tensor(input_data,dtype=torch.float32)
  model.eval()
  with torch.inference_mode():
    logit=model(X)
    pred_prob=torch.sigmoid(logit)
    pred_label=torch.round(pred_prob)
  
  return {"prediction":pred_label.item()}