import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, Trainer, TrainingArguments
from datasets import load_dataset

# Load Dataset
dataset_url = "tahiryaqoob/BISELahore"  # Replace with your dataset repository
dataset = load_dataset(dataset_url, split="train")

# Load Pretrained Model and Tokenizer
model_name = "microsoft/DialoGPT-medium"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Assign Padding Token
if tokenizer.pad_token is None:
    tokenizer.pad_token = tokenizer.eos_token  # Use EOS token as padding token

# Fine-tuning Function
def preprocess_data(example):
    inputs = tokenizer(example['question'], truncation=True, padding="max_length", max_length=128)
    outputs = tokenizer(example['answer'], truncation=True, padding="max_length", max_length=128)
    inputs['labels'] = outputs['input_ids']
    return inputs

# Tokenize Dataset
tokenized_dataset = dataset.map(preprocess_data, batched=True)

# Fine-Tune the Model
training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=1,
    per_device_train_batch_size=2,
    save_steps=500,
    save_total_limit=2,
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_dataset,
)

# Train the Model
trainer.train()

# Save the Fine-Tuned Model
model.save_pretrained("./bise_chatbot_model")
tokenizer.save_pretrained("./bise_chatbot_model")

# Define Chatbot Function
def chatbot_response(user_input):
    inputs = tokenizer.encode(user_input, return_tensors="pt")
    outputs = model.generate(inputs, max_length=100, num_return_sequences=1, do_sample=True)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response

# Create Gradio Interface
iface = gr.Interface(
    fn=chatbot_response,
    inputs="text",
    outputs="text",
    title="BISE Lahore Chatbot",
    description="Ask your questions about BISE Lahore services."
)

iface.launch()