Streamlit App not deploying on Heroku

Summary

I am trying to deploy a ML model using Streamlit and Pycaret on Heroku for the first time.

When I try deploying the app, I get the following error: ModuleNotFoundError: No module named ‘pycaret.internal’

Traceback:

    Traceback:
    File "/app/.heroku/python/lib/python3.6/site-packages/streamlit/ScriptRunner.py", line 322, in _run_script
        exec(code, module.__dict__)
    File "/app/Final.py", line 11, in <module>
        tuned_cat=joblib.load('Cat.pkl')
    File "/app/.heroku/python/lib/python3.6/site-packages/joblib/numpy_pickle.py", line 585, in load
        obj = _unpickle(fobj, filename, mmap_mode)
    File "/app/.heroku/python/lib/python3.6/site-packages/joblib/numpy_pickle.py", line 504, in _unpickle
        obj = unpickler.load()
    File "/app/.heroku/python/lib/python3.6/pickle.py", line 1050, in load
        dispatch[key[0]](self)
    File "/app/.heroku/python/lib/python3.6/pickle.py", line 1338, in load_global
        klass = self.find_class(module, name)
    File "/app/.heroku/python/lib/python3.6/pickle.py", line 1388, in find_class
        __import__(module, level=0)

Final.py:

    import streamlit as st
    import joblib
    
    from pycaret.classification import *
    
    
    tuned_cat=joblib.load('Cat.pkl')
    
    
    import tensorflow as tf
    physical_devices = tf.config.list_physical_devices('CPU') 
    
    
    
    def run():
    
    
        add_selectbox = st.sidebar.selectbox(
        "What would you like to do?",
        ("Online Prediction","Batch Prediction"))
    
        st.sidebar.info('This app is created to predict if the applicant should be granted a loan or not.')
        
    
        st.title("Loan Prediction App")
        
        if add_selectbox == 'Online Prediction':
            
            gender = st.selectbox('Gender',['Female','Male'])
            married = st.selectbox('Married',['No','Yes'])
            depend = st.selectbox('Dependents',['0','1','2','3+'])
            edu = st.selectbox('Education',['Graduate','Not Graduate'])
            self = st.selectbox('Self Employed',['No','Yes'])
            app_inc = st.number_input ('Applicant Income')
            co_inc = st.number_input ('Coapplicant Income')
            amt = st.number_input ('Loan Amount')
            term = st.number_input ('Loan Amount Term')
            credit = st.selectbox('Credit History',['0','1'])
            prop_are = st.selectbox('Property Area',['Rural','Semiurban','Urban'])
    
            output=""
    
            test_df = pd.DataFrame()
            test_df['Gender']= [gender]
            test_df['Married']=[married]
            test_df['Dependents']=[depend] 
            test_df['Education']=[edu]
            test_df['Self_Employed']=[self] 
            test_df['ApplicantIncome']=[app_inc] 
            test_df['CoapplicantIncome']=[co_inc] 
            test_df['LoanAmount']=[amt] 
            test_df['Loan_Amount_Term']=[term] 
            test_df['Credit_History']=[credit] 
            test_df['Property_Area']=[prop_are]      
            
    
            if st.button("Predict"):
                Cat_pred=predict_model(tuned_cat,data=test_df)['Label']
                
                output = Cat_pred.values
                
                if(output==0):
                    text="Rejected"
                    st.error(text)
                    
                elif(output==1):
                    text="Approved"
                    st.success(text)
     
        
        if add_selectbox == 'Batch Prediction':
    
            file_upload = st.file_uploader("Upload excel file for predictions", type=["xlsx"])
    
            if file_upload is not None:
                data = pd.read_excel(file_upload)
                
                st.success('File uploaded successfully!')
                
                Cat_pred=predict_model(tuned_cat,data=data)['Label']
                
                data['Prediction']=Cat_pred
    
                st.write(data)
                st.markdown(get_table_download_link(data), unsafe_allow_html=True)
    
    
    import base64
    
    def get_table_download_link(df):
        """Generates a link allowing the data in a given panda dataframe to be downloaded
        in:  dataframe
        out: href string
        """
        csv = df.to_csv(index=False)
        b64 = base64.b64encode(
            csv.encode()
        ).decode()  # some strings <-> bytes conversions necessary here
        return f'<a href="data:file/csv;base64,{b64}" download="Load_Predictions.csv">Download csv file</a>'
    
    
    
    if __name__ == '__main__':
        run()

Expected behavior:

The app should be deployed on Heroku as it works fine on local computer.

Actual behavior:

Debug info

I have the following dependencies in the requirements.txt file:

pycaret==1.0.0
streamlit==0.58.0

I am using Conda and I had to go with the pycaret version 1.0 because when I chose the latest version the slug size on Heroku would go beyond 500M and it would not deploy. Compiled slug size: 552M is too large (max is 500M).

Link to Github

[jroes]

I was able to reproduce this, here’s the Heroku build output, in case it helps:

https://gist.github.com/jroes/a0c0d2ed7f5fed8efb53eae0bd17ca33

I also tried using the latest version of Streamlit, but couldn’t deploy due to slug size:

https://gist.github.com/jroes/97994971cad86a50f80e32c4494c9610

[randyzwitch]

In this case, since the issue appears to be a slug size issue with Heroku, this is beyond the scope of the Streamlit project