AttributeError: 'HalfBinomialLoss' object has no attribute 'get_init_raw_predictions'

FishWaterUnder commented 4 weeks ago

Chaquopy version

Version 3.9

Devices or emulators where the issue happens

TECHNO KI5K phone

Relevant parts of your code

import numpy as np
import pandas as pd
from os.path import dirname, join
from landmarkers import landmarkers
import pickle
import sklearn

def henllo(lk):
    filename = join(dirname(__file__), "bicep_Curls.pk1")
    with open(filename, 'rb') as f:
        model = pickle.load(f)

    row =[lk.get(i) for i in range(lk.size())]
    X = pd.DataFrame([row], columns = landmarkers)
    bodylang_class = model.predict(X)[0]
    print(bodylang_class)

Describe your issue

I tried to run a machine learning algorithm using a pickle file but it gave me this error

JNI DETECTED ERROR IN APPLICATION: JNI GetObjectClass called with pending exception java.lang.IllegalStateException: com.chaquo.python.PyException: AttributeError: 'HalfBinomialLoss' object has no attribute 'get_init_raw_predictions'

I followed this person in #160 but theirs work I believe when my doesn't

mhsmith commented 3 weeks ago

If you search for this error, you'll find https://github.com/scikit-learn/scikit-learn/issues/14076, which says:

While models saved using one version of scikit-learn might load in other versions, this is entirely unsupported and inadvisable.

So you'll have to either:

Regenerate the model with the same version of scikit-learn as you're using with your app; or
Make your app use a different scikit-learn version which is closer to the version that generated the model. Unfortunately there are only a few versions available.

mhsmith commented 3 weeks ago

There may still be a Chaquopy bug here, because it should have produced a regular Java exception rather than a JNI error. But I have no other reports of this happening, so in order to look into it, I would need a complete and self-contained example.

FishWaterUnder commented 2 weeks ago

Thank You! It worked

chaquo / chaquopy