import torch
from transformers import T5ForConditionalGeneration, T5TokenizerFast
import gradio as gr

# ============================================================
# CONFIG
# ============================================================
MODEL_NAME = "t5-small"
WEIGHTS_PATH = "t5_weights.pt"  # your uploaded weights file

# ============================================================
# LOAD MODEL & TOKENIZER
# ============================================================
device = "cuda" if torch.cuda.is_available() else "cpu"

tokenizer = T5TokenizerFast.from_pretrained(MODEL_NAME)
model = T5ForConditionalGeneration.from_pretrained(MODEL_NAME)
model.load_state_dict(torch.load(WEIGHTS_PATH, map_location=device))
model.to(device)
model.eval()

# ============================================================
# SUMMARIZATION FUNCTION
# ============================================================
def summarize(text):
    if not text.strip():
        return "⚠️ Please enter some text."
    inputs = tokenizer(
        "summarize: " + text,
        return_tensors="pt",
        truncation=True,
        max_length=512,
        padding="max_length"
    ).to(device)

    summary_ids = model.generate(
        **inputs,
        max_new_tokens=150,
        num_beams=4,
        early_stopping=True,
        no_repeat_ngram_size=3
    )

    return tokenizer.decode(summary_ids[0], skip_special_tokens=True)

# ============================================================
# GRADIO UI
# ============================================================
demo = gr.Interface(
    fn=summarize,
    inputs=gr.Textbox(lines=10, label="Enter News Article"),
    outputs=gr.Textbox(lines=8, label="Generated Summary"),
    title="📰 T5 News Summarizer",
    description="Upload your own fine-tuned T5 model weights and summarize news articles easily!",
    examples=[
        ["The US economy grew at an annual rate of 2.4% last quarter, surprising analysts..."],
        ["Scientists have discovered a new species of bird in the Amazon rainforest..."]
    ]
)

# ============================================================
# RUN APP
# ============================================================
if __name__ == "__main__":
    demo.launch()