Docker image doesn't run with XGBoost version 1.5.0

#1

I trained an XGBoost model(version 1.5.0) for regression problem.
Then I successfully built a docker and I tried to run it but nothing happens:

Deployment docker run --rm -p 80:80 gehalt-bench:no-batch

INFO:     Started server process [1]
INFO:     Waiting for application startup

So after “Waiting for application startup” nothing happens.

Here is my code for server file:

import pickle
import numpy as np
from fastapi import FastAPI
from pydantic import BaseModel


app = FastAPI(title="Salary")

class Salary(BaseModel):
    attr1: float
    attr2: int
    attr3: int
    attr4: int 
    attr5: int
    attr6: int
    attr7: int
    attr8: int

    

#function run at the startup of the server
@app.on_event('startup')
def load_model():
    with open('/app/model.pkl','rb') as file:
        global model
        model = pickle.load(file)


@app.post('/predict')
def predict(sal:Salary):
    data_point = np.array(
        [
            [
                sal.attr1,
                sal.attr2,
                sal.attr3,
                sal.attr4,
                sal.attr5,
                sal.attr6,
                sal.attr7,
                sal.attr8
            ]
        ]
    )

    pred = model.predict(data_point).tolist()
    pred = pred[0]
    print(pred)
    return {'Prediction':pred}

my requirements.txt file:

fastapi
uvicorn
scikit-learn==0.24.1
xgboost==1.5.0

Docker File:

FROM frolvlad/alpine-miniconda3:python3.7

COPY requirements.txt .

RUN pip install -r requirements.txt && \ 
        rm requirements.txt

EXPOSE 80

COPY ./app /app

CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "80"]

How can I fix a problem? It might be, that version of xgboost might be a problem…
WIth other sklearn models everything works fine.

#2

You should install libgomp package, since XGBoost depends on it.

1 Like