app.py

import torch
import gradio as gr
from modeling_diffusion import DiffusionTextModel

# =====================
# Load Model from Hub
# =====================
model = DiffusionTextModel.from_pretrained("yasserrmd/diffusion-text-demo")
model.eval()

device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

PAD_TOKEN = "[PAD]"
MASK_TOKEN = "[MASK]"

vocab = {PAD_TOKEN: 0, MASK_TOKEN: 1}

# Reverse mapping
id_to_word = {i: w for w, i in vocab.items()}

# Special token IDs
pad_id = vocab[PAD_TOKEN]
mask_id = vocab[MASK_TOKEN]



# =====================
# Generation Function
# =====================
def generate_with_prompt(model, input_text, max_length=50, T=10):
    # Ensure max_length does not exceed 99
    max_length = min(max_length, 99)

    model.eval()
    input_tokens = input_text.split()
    input_ids = [vocab.get(tok, mask_id) for tok in input_tokens]

    seq = torch.full((1, max_length), mask_id, dtype=torch.long, device=device)
    seq[0, :len(input_ids)] = torch.tensor(input_ids, device=device)

    for step in range(T, 0, -1):
        with torch.no_grad():
            logits = model(seq, torch.tensor([step], device=device))
            probs = torch.softmax(logits, dim=-1)
            for pos in range(len(input_ids), max_length):
                if seq[0, pos].item() == mask_id:
                    seq[0, pos] = torch.multinomial(probs[0, pos], 1)

    ids = seq[0].tolist()
    if pad_id in ids:
        ids = ids[:ids.index(pad_id)]
    return " ".join(id_to_word[i] for i in ids)

# =====================
# Gradio App
# =====================
def chat_fn(message, history, steps, max_len):
    response = generate_with_prompt(model, message, max_length=max_len, T=steps)
    history.append((message, response))
    return "", history

with gr.Blocks() as demo:
    gr.Markdown("## 🌀 DiffusionTextModel QA Chat Demo")
    chatbot = gr.Chatbot()
    msg = gr.Textbox(placeholder="Type your question or prompt here...")
    steps = gr.Slider(1, 50, value=10, step=1, label="Diffusion Steps (T)")
    max_len = gr.Slider(10, 99, value=50, step=1, label="Max Token Length (≤ 99)")
    clear = gr.Button("Clear")

    msg.submit(chat_fn, [msg, chatbot, steps, max_len], [msg, chatbot])
    clear.click(lambda: None, None, chatbot, queue=False)

demo.launch()