I’m trying to use fastapi e uvicorn to make a prediction. I’ve already trained the model on my Pycharm project. It works. Next step is to deploy via FastAPI. Should be kinda straightforward: in the url/docs i should load the pickle file with the weights and the csv with the data. Then i should retrieve the output with the results.

I wrote this:

import io
from fastapi import FastAPI, UploadFile, File
import pandas as pd
import joblib

import uvicorn


# instance
app = FastAPI(title="Breast cancer ML API", description="API for classification of breast cancer", version="1.0", predix="asd", tags=["items"])


@app.post("/predict/")
async def predict(file: UploadFile = File(...), weights: UploadFile = File(...)):
    content = await weights.read()
    model = joblib.load(content)
    test = await file.read()
    df = pd.read_csv(io.StringIO(test.decode('utf-8')))
    preprocessing = model.named_steps("preprocessing")
    preprocessed_data = preprocessing.transform(df)
    # prediction
    classifier = model.named_steps("classifier")
    prediction = classifier.predict(preprocessed_data)

    prediction_df = pd.DataFrame(prediction, index=df.index, columns=["diagnosis"])
    result = prediction_df.to_csv()
    return len(df)


if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=8000)

but i have UniCodeDecodeError. So, how do i upload and read the csv as pandas dataframe, in order to be preprocessed and classified as my model?

And, how do i read the pickle correctly?

Thanks for answers!