model.py

import json
import torch
from transformers import BertTokenizer, BertForSequenceClassification, Trainer, TrainingArguments
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import torch.nn as nn
from torch.utils.data import Dataset

# Load the data from intents.json
with open("data/intents.json") as file:
    intents_data = json.load(file)

# Initialize the tokenizer and model
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = BertForSequenceClassification.from_pretrained("bert-base-uncased", num_labels=len(set([intent["tag"] for intent in intents_data["intents"]])))

# Prepare the data: tokenize and encode the text
train_data = []
train_labels = []

for intent in intents_data["intents"]:
    for pattern in intent["patterns"]:
        # Tokenize the input text
        encoded_input = tokenizer(pattern, padding=True, truncation=True, return_tensors="pt")
        train_data.append(encoded_input)
        train_labels.append(intent["tag"])

# Encode the labels (e.g., "greeting", "goodbye") to numeric values
label_encoder = LabelEncoder()
train_labels_encoded = label_encoder.fit_transform(train_labels)

# Split the data into training and testing sets
train_data, test_data, train_labels, test_labels = train_test_split(train_data, train_labels_encoded, test_size=0.2)

# Create a custom dataset class for PyTorch
class ChatbotDataset(Dataset):
    def __init__(self, data, labels):
        self.data = data
        self.labels = labels

    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        return {
            'input_ids': self.data[idx]['input_ids'].squeeze(),
            'attention_mask': self.data[idx]['attention_mask'].squeeze(),
            'labels': torch.tensor(self.labels[idx])
        }

train_dataset = ChatbotDataset(train_data, train_labels)
test_dataset = ChatbotDataset(test_data, test_labels)

# Training setup
training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=3,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    logging_dir="./logs",
    evaluation_strategy="epoch",  # Evaluate at the end of each epoch
    save_strategy="epoch",       # Save the model at the end of each epoch
)

# Initialize the Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
    eval_dataset=test_dataset,
)

# Train the model
trainer.train()

# Save the trained model and tokenizer
model.save_pretrained("./results")
tokenizer.save_pretrained("./results")

# Save the label encoder for future inference
import pickle
with open('label_encoder.pkl', 'wb') as f:
    pickle.dump(label_encoder, f)

print("Training complete. Model and tokenizer saved.")