pipeline.py

import torch
import joblib
import librosa
import numpy as np
from torch import nn
from transformers import AutoModel

class VoiceRecognitionModel(nn.Module):
    def __init__(self, num_classes):
        super().__init__()
        # Your model architecture here (same as training)
        self.conv1 = nn.Conv2d(1, 32, kernel_size=3, padding=1)
        # ... rest of your architecture
        
    def forward(self, x):
        # Your forward pass
        return x

def extract_features(file_path, max_pad_len=174):
    # Your feature extraction code
    pass

def pipeline():
    # This will be called when someone uses your model
    model = VoiceRecognitionModel(num_classes=7)  # Adjust based on your classes
    model.load_state_dict(torch.load("voice_recognition_model.pth"))
    model.eval()
    
    label_encoder = joblib.load("label_encoder.joblib")
    feature_params = joblib.load("feature_params.joblib")
    
    return model, label_encoder, feature_params