app.py

"""
Gradio app for SmolLM2-135M inference with streaming output.
Loads model from Hugging Face model repo.
"""

import sys
from pathlib import Path
from typing import List, Optional
import os

import gradio as gr
import torch
from transformers import AutoConfig, AutoTokenizer
from huggingface_hub import hf_hub_download

from model import SmolConfig, SmolLM2
from train import SmolLM2Module

# Hugging Face model repo configuration
HF_MODEL_REPO = "Sualeh77/smollm2-135m-trained-on-tinyShakespear-forfun"
CHECKPOINT_NAME = "smollm2-step=05000-train_loss=0.0918.ckpt"

# Device setup
DEVICE = "cpu"
if torch.cuda.is_available():
    DEVICE = "cuda"
elif hasattr(torch.backends, "mps") and torch.backends.mps.is_available():
    DEVICE = "mps"

# Globals
model: Optional[SmolLM2] = None
tokenizer = None

# Allow SmolConfig to be deserialized from Lightning checkpoints when torch.load
try:
    torch.serialization.add_safe_globals([SmolConfig])  # type: ignore[attr-defined]
except Exception:
    pass


def load_model_checkpoint(checkpoint_path: Optional[str] = None, use_hf: bool = True):
    """Load Lightning checkpoint from Hugging Face Hub or local path."""
    global model, tokenizer

    try:
        # Load tokenizer and config from Hugging Face
        hf_cfg = AutoConfig.from_pretrained("HuggingFaceTB/SmolLM2-135M")
        config = SmolConfig.from_hf(hf_cfg)
        tokenizer = AutoTokenizer.from_pretrained("HuggingFaceTB/SmolLM2-135M")
        if tokenizer.pad_token is None:
            tokenizer.pad_token = tokenizer.eos_token

        # Determine checkpoint path
        if use_hf and checkpoint_path is None:
            # Download from Hugging Face Hub
            try:
                local_ckpt = hf_hub_download(
                    repo_id=HF_MODEL_REPO,
                    filename=CHECKPOINT_NAME,
                    cache_dir=None,  # Use default cache
                )
                checkpoint_path = local_ckpt
                status_msg = f"✅ Model loaded from Hugging Face: {HF_MODEL_REPO}/{CHECKPOINT_NAME}"
            except Exception as e:
                return f"❌ Failed to download from HF Hub: {e}"
        elif checkpoint_path:
            # Use local path
            ckpt = Path(checkpoint_path)
            if not ckpt.exists():
                return f"❌ Checkpoint not found: {ckpt}"
            status_msg = f"✅ Model loaded from local path: {checkpoint_path}"
        else:
            return "❌ No checkpoint path provided"

        # Load the Lightning module
        module = SmolLM2Module.load_from_checkpoint(
            str(checkpoint_path),
            config=config,
            tokenizer=tokenizer,
            map_location=DEVICE,
            strict=False,
        )
        module.eval()
        model = module.model.to(DEVICE).eval()
        return f"{status_msg} on {DEVICE}"
    except Exception as e:
        model = None
        return f"❌ Error loading model: {e}"


def stream_generate(
    prompt: str,
    max_new_tokens: int,
    temperature: float,
    top_k: int,
    top_p: float,
):
    """Generator that yields only the generated text (without prompt)."""
    global model, tokenizer
    if model is None or tokenizer is None:
        yield "⚠️ Load the model first (click Reload Model)."
        return

    if not prompt or not prompt.strip():
        yield "⚠️ Please enter a prompt."
        return

    # Tokenize prompt
    inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
    input_ids = inputs["input_ids"].to(DEVICE)

    # Guard against context overflow
    if input_ids.shape[1] >= model.config.max_position_embeddings:
        yield f"⚠️ Prompt too long ({input_ids.shape[1]} tokens). Max is {model.config.max_position_embeddings}."
        return

    generated = input_ids
    past_key_values: Optional[List] = None
    prompt_length = input_ids.shape[1]

    with torch.no_grad():
        for _ in range(max_new_tokens):
            if past_key_values is None:
                current_input = generated
            else:
                current_input = generated[:, -1:]

            logits, past_key_values = model(
                current_input,
                past_key_values=past_key_values,
                use_cache=True,
            )

            next_token_logits = logits[:, -1, :] / max(temperature, 1e-6)

            # top-k
            if top_k > 0:
                values, _ = torch.topk(next_token_logits, top_k)
                min_keep = values[:, -1].unsqueeze(-1)
                next_token_logits = torch.where(
                    next_token_logits < min_keep,
                    torch.full_like(next_token_logits, float("-inf")),
                    next_token_logits,
                )

            # top-p
            if top_p < 1.0:
                sorted_logits, sorted_indices = torch.sort(next_token_logits, descending=True)
                probs = torch.softmax(sorted_logits, dim=-1)
                cumulative = torch.cumsum(probs, dim=-1)
                sorted_mask = cumulative > top_p
                sorted_mask[..., 1:] = sorted_mask[..., :-1].clone()
                sorted_mask[..., 0] = 0
                mask = sorted_mask.scatter(1, sorted_indices, sorted_mask)
                next_token_logits = torch.where(mask, torch.full_like(next_token_logits, float("-inf")), next_token_logits)

            probs = torch.softmax(next_token_logits, dim=-1)
            next_token = torch.multinomial(probs, num_samples=1)

            generated = torch.cat([generated, next_token], dim=1)
            # Decode only the generated part (skip the prompt)
            generated_text = tokenizer.decode(generated[0][prompt_length:], skip_special_tokens=True)
            yield generated_text


# Initial load from Hugging Face
INITIAL_STATUS = load_model_checkpoint(use_hf=True)


def chat_stream(message, history, max_tokens, temperature, top_k, top_p):
    """Gradio wrapper for streaming chat."""
    if history is None:
        history = []

    # Convert history from tuple format to dict format if needed
    if history and isinstance(history[0], (list, tuple)):
        new_history = []
        for h in history:
            if isinstance(h, (list, tuple)) and len(h) >= 2:
                if h[0]:  # User message
                    new_history.append({"role": "user", "content": str(h[0])})
                if h[1]:  # Assistant message
                    new_history.append({"role": "assistant", "content": str(h[1])})
        history = new_history

    # Append user message
    user_msg = (message or "").strip()
    if not user_msg:
        yield history
        return
    
    history.append({"role": "user", "content": user_msg})
    history.append({"role": "assistant", "content": ""})

    stream = stream_generate(user_msg, max_tokens, temperature, top_k, top_p)
    for partial in stream:
        # Update the last assistant message with generated text
        if partial:
            history[-1] = {"role": "assistant", "content": str(partial)}
        yield history


def clear_chat():
    return "", []


with gr.Blocks(title="SmolLM2-135M Text Generator") as demo:
    gr.Markdown(
        """
        # 🤖 SmolLM2-135M Text Generator

        Generate text with your trained SmolLM2-135M model (streaming output).
        
        **Model:** Trained on TinyShakespeare dataset
        **Source:** [Hugging Face Model Repo](https://huggingface.co/Sualeh77/smollm2-135m-trained-on-tinyShakespear-forfun)
        """
    )

    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("### Model Status")
            status_text = gr.Textbox(value=INITIAL_STATUS, label="Status", interactive=False, lines=3)
            load_btn = gr.Button("🔄 Reload Model from HF", variant="secondary")
            load_btn.click(fn=lambda: load_model_checkpoint(use_hf=True), outputs=status_text)
            
            gr.Markdown("### Local Checkpoint (Optional)")
            ckpt_input = gr.Textbox(
                value="",
                label="Local checkpoint path (leave empty to use HF)",
                interactive=True,
            )
            load_local_btn = gr.Button("📁 Load from Local Path", variant="secondary")
            load_local_btn.click(
                fn=lambda p: load_model_checkpoint(checkpoint_path=p, use_hf=False) if p else "⚠️ Please enter a path",
                inputs=ckpt_input,
                outputs=status_text
            )

            gr.Markdown("### Generation Parameters")
            max_tokens = gr.Slider(10, 500, value=100, step=10, label="Max Tokens")
            temperature = gr.Slider(0.1, 2.0, value=0.8, step=0.1, label="Temperature")
            top_k = gr.Slider(0, 100, value=50, step=5, label="Top-K")
            top_p = gr.Slider(0.1, 1.0, value=1.0, step=0.05, label="Top-P")

        with gr.Column(scale=2):
            gr.Markdown("### 💬 Chat Interface")
            chatbot = gr.Chatbot(label="Conversation", height=500)
            with gr.Row():
                msg = gr.Textbox(label="Your Message", placeholder="Type your prompt here...", scale=4, lines=2)
                submit_btn = gr.Button("Send ➤", variant="primary", scale=1)
            clear_btn = gr.Button("🗑️ Clear Chat", variant="stop")

    msg.submit(fn=chat_stream, inputs=[msg, chatbot, max_tokens, temperature, top_k, top_p], outputs=chatbot)
    submit_btn.click(fn=chat_stream, inputs=[msg, chatbot, max_tokens, temperature, top_k, top_p], outputs=chatbot).then(fn=lambda: "", outputs=msg)
    clear_btn.click(fn=clear_chat, outputs=[msg, chatbot])


if __name__ == "__main__":
    demo.queue().launch(share=False, server_name="0.0.0.0", server_port=7860)