training.py

"""
training.py
Fonctions d'entraînement pour sklearn et Keras, sauvegarde en bytes pour téléchargement.
"""
from sklearn.base import clone
from sklearn.metrics import accuracy_score
import joblib
import io
import tempfile
import numpy as np
import streamlit as st
from tensorflow.keras.callbacks import EarlyStopping

def train_sklearn_model(model, X_train, y_train, X_test=None, y_test=None):
    """
    Entraîne un modèle sklearn et retourne (model, history_dict)
    """
    model.fit(X_train, y_train)
    history = {}
    if X_test is not None and y_test is not None:
        try:
            preds = model.predict(X_test)
            acc = accuracy_score(y_test, preds)
            history['test_accuracy'] = float(acc)
        except Exception:
            pass
    return model, history

def train_keras_model(model, X_train, y_train, X_val, y_val, epochs=10, batch_size=32):
    cb = [EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)]
    history = model.fit(X_train, y_train, validation_data=(X_val, y_val),
                        epochs=epochs, batch_size=batch_size, callbacks=cb, verbose=0)
    return history.history, model

def save_model_to_bytes(model, model_type="sklearn"):
    """
    Retourne un buffer bytes prêt pour st.download_button
    """
    buffer = io.BytesIO()
    if model_type == "sklearn":
        joblib.dump(model, buffer)
        buffer.seek(0)
    elif model_type == "keras":
        # save model to temporary file then read bytes
        import tempfile
        tmp = tempfile.NamedTemporaryFile(suffix=".h5", delete=False)
        model.save(tmp.name)
        tmp.close()
        with open(tmp.name, "rb") as f:
            buffer.write(f.read())
        buffer.seek(0)
    else:
        raise ValueError("model_type must be 'sklearn' or 'keras'")
    return buffer.getvalue()