我正在尝试学习如何使用 FastAPI 创建 API。为此，我决定使用 Streamlight python 包创建一个应用程序，并调用 API 以提供基于我使用 TensorFlow 准备的机器学习模型的响应。
问题是，当我通过“.post”调用 API 时，它会返回错误 422，我已经测试了多种方法来解决这个问题，但最终导致 API 崩溃或无法工作。
API 目前部署在 Heroku (https://heart-api-first.herokuapp.com/docs) 上
代码如下：
Streamlit 应用（前端）：
# Loading packages ##########################################################################
import streamlit as st
import streamlit.components.v1 as components
import numpy as np
import pandas as pd
import requests,json
###############################################################################################
st.set_page_config(layout="wide")
# Loading of the data
@st.cache
def load_data():
data=pd.read_csv('heart_failure_clinical_records_dataset.csv')
return data
df = load_data()
# Sidebar creation ####################################################################
st.sidebar.title('Patient Parameters')
# Creating parameters
params = {
'Age':st.sidebar.slider('Age', min(df['age']), max(df['age']),1.0),
'Anemia':st.sidebar.selectbox('Anemia', ('Yes','No')),
'CPK':st.sidebar.slider('CPK [mcg/L]', min(df['creatinine_phosphokinase']), max(df['creatinine_phosphokinase']), 1),
'Diabetes':st.sidebar.selectbox('Diabetes', ('Yes','No')),
'Ejection fraction':st.sidebar.slider('Ejection fraction [%]', min(df['ejection_fraction']), max(df['ejection_fraction']), 1),
'High Blood Pressure':st.sidebar.selectbox('High Blood Pressure', ('Yes','No')),
'Platelets in the blood':st.sidebar.slider('Platelets in the blood [kiloplatelets/mL]', min(df['platelets']), max(df['platelets']),1.0),
'Serum Creatinine':st.sidebar.slider('Serum Creatinine [mg/dL]', min(df['serum_creatinine']), max(df['serum_creatinine']),1.0),
'Serum Sodium':st.sidebar.slider('Serum Sodium [mEq/L]', min(df['serum_sodium']), max(df['serum_sodium']),1),
'Woman or Man':st.sidebar.selectbox('Woman or Man', ('Woman','Men')),
'Smoking':st.sidebar.selectbox('Smoking',('Yes','No')),
'Time':st.sidebar.slider('Time [days]', min(df['time']), max(df['time']),1)
}
# Method for changing attribute
def change_type(X):
if X == 'Yes':
X = 1
else:
X = 0
return X
def change_sex(X):
if X == 'Woman':
X = 1
else:
X = 0
return X
# Defining user input
def user_input():
age = params['Age']
anemia = params['Anemia']
cpk = params['CPK']
diabetes = params['Diabetes']
ejection_fraction = params['Ejection fraction']
hbp = params['High Blood Pressure']
platelets = params['Platelets in the blood']
serum_creatinine = params['Serum Creatinine']
serum_sodium = params['Serum Sodium']
woman_man = params['Woman or Man']
smoking = params['Smoking']
time = params['Time']
data = {
'Age': age,
'Anemia': change_type(anemia),
'CPK': cpk,
'Diabetes':change_type(diabetes),
'Ejection_fraction':ejection_fraction,
'HBP':change_type(hbp),
'Platelets':platelets,
'Serum_Creatinine':serum_creatinine,
'Serum_Sodium':serum_sodium,
'Woman_Man':change_sex(woman_man),
'Smoking':change_type(smoking),
'Time':time
}
# features = pd.DataFrame(data, index=[0])
return data
# API information
url = 'https://heart-api-first.herokuapp.com'
endpoint = '/prediction'
# defining process for calling API
def process(data, server_url: str):
heart_data = data
result = requests.post(server_url,data = heart_data)
return result
# Expander ##################################################################################
with st.beta_expander('Information', expanded=True):
st.header('Goal of the Web Application')
st.write('The main goal of this project to learn on how to use the fastAPI package for creating background services, in this case, a machine learning model.')
st.write('**Source of data**: https://www.kaggle.com/andrewmvd/heart-failure-clinical-data')
st.write("""
> **Attention**:
* This app is **NOT** meant as a medical device/applicaiton for medical diagnosis;
* It is just for educational purpose;
* If you think you have an health problem, please reach out to an authorized medical professional;
* The creator of this applicaiton will **NOT** assume any responsability for any misuse of the application.
""")
st.header('Parameters Information')
st.write("""
* **Age**: Age of the patient;
* **Anemia**: Decrease of red blood cells or hemoglobin;
* **Creatinine Phosphokinase Enzyme (CPK)**: Level of CPK enzyme in the blood (mcg/L);
* **Diabetes**: Patient is diabetic or not;
* **Ejection Fraction**: Percentage of blood that leaves the heart at each contraction;
* **High Blood Pressure**: Patient has high blood pressure;
* **Platelets in the blood**: Number of platelets in the blood [kiloplatelets/mL];
* **Serum Creatinine**: Level of serum creatinine in the blood [mg/dL];
* **Serum Sodium**: Level of serum sodium in the blood [mEq/L];
* **Woman or Man**: If patient is a woman or man;
* **Smoking**: If patient smokes or not;
* **Time**: Follow-up periods, number of days after the first check.
""")
# Columns creation #####################################################################
left_column1, right_column1 = st.beta_columns(2)
with left_column1:
st.header('Heart Animation')
st.write('In this GIF, you can see how the heart works, with the name of its major components.')
components.iframe("https://healthblog.uofmhealth.org/sites/consumer/files/2020-01/heart_pumping.gif",height=500)
with right_column1:
st.header('Heart Attack Video')
st.write('In this video, created by the YouTube channell _Institute of Human Anatomy_, explain what happens during a heart attack.')
st.write('**Attention**: Video contains content (real human heart from a dead body) that may cause some problem for some people.')
st.video("https://www.youtube.com/watch?v=m0fiHGlqivo&t=1s")
# Creating the data ####################################################################
user_input_original_df = user_input()
user_input_original_df
# Creating button
test_df = pd.DataFrame(user_input_original_df).dtypes
test_df
# if st.sidebar.button("Predict"):
# result = process(user_input_original_df, url+endpoint)
# result
# else:
# pass
API：
import uvicorn
from fastapi import FastAPI
# Download basic ML model for creating class
from pydantic import BaseModel
# Code for creating the model of the API
import numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import OneHotEncoder
# Neural Netwrok libraries
import tensorflow as tf
from tensorflow import keras
#from tensorflow.keras import Sequential
#from tensorflow.keras.layers import Dense
# Creating class for validation
class Heart(BaseModel):
Age: float
Anemia: int
CPK: int
Diabetes: int
Ejection_fraction: int
HBP: int
Platelets: float
Serum_Creatinine: float
Serum_Sodium: int
Woman_Man: int
Time: int
# Creating std scaler
sc = StandardScaler()
# Loading the model saved
model = keras.models.load_model("Test_model_Neural_Network")
# List for prediction
pred_label = list()
# Defining data preparation
def preparazione(dati_originali):
dati_pronti = dati_originali.values
dati_pronti = sc.fit_transform(dati_originali)
return dati_pronti
# Defining function for prediction
def predict(dati):
previsione = model.predict(dati)
return previsione
# Defining method for converting results
def convert(dati_preparati, lista):
for i in range(len(dati_preparati)):
result = pred_label.append(np.argmax(dati_preparati[i]))
return result
app = FastAPI()
@app.get("/")
def home():
return {"message":"Welcome to my first API!"}
@app.post("/prediction")
async def get_prediction(heart :Heart):
input_df = pd.DataFrame([heart.dict()])
dati_preparati = preparazione(input_df)
prediction_heart = predict(dati_preparati)
prediction_label = convert(prediction_heart, pred_label)
return prediction_label
if __name__ == '__main__':
uvicorn.run(app, host='127.0.0.1', port=8000)
如果您对如何解决此问题有任何建议，我将不胜感激。