fine_tune_bert.py

# fine_tune_bert.py
from transformers import BertTokenizer, BertForSequenceClassification, Trainer, TrainingArguments
from datasets import load_dataset

# Assuming you have preprocessed ChatLog.txt into a suitable format and loaded it as a dataset
dataset = load_dataset('csv', data_files={'train': 'chat_data.csv'})

# Initialize the tokenizer and model for Bert
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model = BertForSequenceClassification.from_pretrained('bert-base-uncased')

# Tokenize the input text
def tokenize_function(examples):
    return tokenizer(examples["text"], padding="max_length", truncation=True)

tokenized_datasets = dataset.map(tokenize_function, batched=True)

# Define training arguments
training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=3,
    per_device_train_batch_size=8,
    warmup_steps=500,
    weight_decay=0.01,
    logging_dir="./logs",
)

# Initialize Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets["train"],
)

# Train the model
trainer.train()

# Save the fine-tuned model
model.save_pretrained('./fine_tuned_model')