# Model Loader Script - Compatible with Python 3.13.9 import torch import torch.nn as nn from transformers import AutoModel, AutoTokenizer import json import os class MultiHeadClassifier(nn.Module): def __init__(self, bert_model_name="aubmindlab/bert-base-arabertv2", num_heads=5, num_classes=4): super().__init__() self.bert = AutoModel.from_pretrained(bert_model_name) hidden_size = self.bert.config.hidden_size self.heads = nn.ModuleList([ nn.Sequential( nn.Linear(hidden_size, 256), nn.ReLU(), nn.Dropout(0.3), nn.Linear(256, num_classes) ) for _ in range(num_heads) ]) def forward(self, input_ids, attention_mask): outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask) cls_embedding = outputs.last_hidden_state[:, 0, :] logits = [head(cls_embedding) for head in self.heads] return logits def load_feedback_model(model_path, tokenizer_path): """Load the trained model for inference""" # Load config with open(os.path.join(os.path.dirname(model_path), "model_config.json"), "r") as f: config = json.load(f) # Initialize model model = MultiHeadClassifier( bert_model_name=config["bert_model_name"], num_heads=config["num_heads"], num_classes=config["num_classes"] ) # Load weights model.load_state_dict(torch.load(model_path, map_location="cpu")) model.eval() # Load tokenizer tokenizer = AutoTokenizer.from_pretrained(tokenizer_path) return model, tokenizer, config # Example usage: # model, tokenizer, config = load_feedback_model( # "saved_model_python313/feedback_classifier.pt", # "saved_model_python313/tokenizer" # )