src/play_interface.py

"""Interface to play against the model.
"""

import huggingface_hub
import chess
import chess.svg
import uuid
import random
import wandb

import gradio as gr

from . import constants

model_name = "yp-edu/gpt2-stockfish-debug"
headers = {"X-Wait-For-Model": "true"}
client = huggingface_hub.InferenceClient(
    model=model_name, headers=headers
)
inference_fn = client.text_generation


def plot_board(
    board: chess.Board,
):
    try:
        last_move = board.peek()
        arrows = [(last_move.from_square, last_move.to_square)]
    except IndexError:
        arrows = []
    if board.is_check():
        check = board.king(board.turn)
    else:
        check = None
    svg_board = chess.svg.board(
        board,
        check=check,
        size=350,
        arrows=arrows,
    )
    id = str(uuid.uuid4())
    with open(f"{constants.FIGURE_DIRECTORY}/board_{id}.svg", "w") as f:
        f.write(svg_board)
    return f"{constants.FIGURE_DIRECTORY}/board_{id}.svg"

def render_board(
    current_board: chess.Board,
):
    fen = current_board.fen()
    pgn = current_board.root().variation_san(current_board.move_stack)
    image_board = plot_board(current_board)
    return fen, pgn, image_board

def play_user_move(
    uci_move: str,
    current_board: chess.Board,
):
    current_board.push_uci(uci_move)
    return current_board

def play_ai_move(
    current_board: chess.Board,
    temperature: float = 0.1,
    top_k: int = 3,
):
    uci_move = inference_fn(
        prompt=f"FEN: {current_board.fen()}\nMOVE:",
        temperature=temperature,
        top_k=top_k,
    )
    current_board.push_uci(uci_move.strip())
    return current_board

def try_play_move(
    username: str,
    move_to_play: str,
    current_board: chess.Board,
):
    if current_board.is_game_over():
        gr.Warning("The game is already over")
        return *render_board(current_board), current_board
    try:
        current_board = play_user_move(move_to_play, current_board)
        if current_board.is_game_over():
            gr.Info(f"Congratulations, {username}!")
            with wandb.init(project="gpt2-stockfish-debug", entity="yp-edu") as run:
                run.log(
                    {
                        "username": username,
                        "winin": current_board.fullmove_number,
                        "pgn": current_board.variation_san(current_board.move_stack),
                    }
                )
                run.finish()
            return *render_board(current_board), current_board
    except:
        gr.Warning("Invalid move")
        return *render_board(current_board), current_board
    temperature_retries = [
        (i+1)/10 for i in range(10)
    ]
    for temperature in temperature_retries:
        try:
            current_board = play_ai_move(current_board, temperature=temperature)
            break
        except:
            gr.Warning(f"AI move failed with temperature {temperature}")
    else:
        gr.Warning("AI move failed with all temperatures")
        current_board.pop()
        return *render_board(current_board), current_board


with gr.Blocks() as interface:
    with gr.Column():
        username = gr.Textbox(
            label="Username to record on leaderboard (should you win)",
            lines=1,
            max_lines=1,
            value="",
        )
        current_fen = gr.Textbox(
            label="Board FEN",
            lines=1,
            max_lines=1,
            value=chess.STARTING_FEN,
        )
        current_pgn = gr.Textbox(
            label="Action sequence",
            lines=1,
            value="",
        )
        with gr.Row():
            move_to_play = gr.Textbox(
                label="Move to play (UCI)",
                lines=1,
                max_lines=1,
                value="",
            )
            play_button = gr.Button("Play")

    with gr.Column():
        image_board = gr.Image(label="Board")

    static_inputs = [
        username,
        move_to_play,
    ]
    static_outputs = [
        current_fen,
        current_pgn,
        image_board,
    ]
    is_ai_white = random.choice([True, False])
    init_board = chess.Board()
    if is_ai_white:
        init_board = play_ai_move(init_board)
    state_board = gr.State(value=init_board)
    play_button.click(
        try_play_move,
        inputs=[*static_inputs, state_board],
        outputs=[*static_outputs, gr.State()],
    )
    interface.load(render_board, inputs=[state_board], outputs=[*static_outputs])