Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,6 +4,7 @@ from llama_cpp import Llama
|
|
| 4 |
from huggingface_hub import hf_hub_download
|
| 5 |
import numpy as np
|
| 6 |
from typing import List
|
|
|
|
| 7 |
|
| 8 |
model = Llama(
|
| 9 |
model_path=hf_hub_download(
|
|
@@ -47,6 +48,16 @@ def extract_xml_move(text: str) -> str:
|
|
| 47 |
return match.group(1)
|
| 48 |
return ""
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
def convert_moves_to_coordinate_list(moves_list: List[str]) -> str:
|
| 51 |
"""
|
| 52 |
Converts a list of moves to a coordinate list representation.
|
|
@@ -291,6 +302,9 @@ def create_interface():
|
|
| 291 |
margin: 15px 0;
|
| 292 |
font-family: monospace;
|
| 293 |
min-height: 100px;
|
|
|
|
|
|
|
|
|
|
| 294 |
}
|
| 295 |
.reasoning-box {
|
| 296 |
border-left: 4px solid #2196F3;
|
|
@@ -313,6 +327,10 @@ def create_interface():
|
|
| 313 |
div.svelte-1nguped {
|
| 314 |
display: block;
|
| 315 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 316 |
"""
|
| 317 |
|
| 318 |
with gr.Blocks(css=css) as interface:
|
|
@@ -352,7 +370,7 @@ def create_interface():
|
|
| 352 |
info="Lower values make AI more deterministic, higher values more creative"
|
| 353 |
)
|
| 354 |
|
| 355 |
-
def
|
| 356 |
if game.game_over:
|
| 357 |
return [
|
| 358 |
render_board(game.board),
|
|
@@ -386,114 +404,145 @@ def create_interface():
|
|
| 386 |
game_state = game.format_game_state()
|
| 387 |
print(game_state)
|
| 388 |
|
| 389 |
-
#
|
| 390 |
-
|
| 391 |
-
|
| 392 |
-
|
| 393 |
-
|
| 394 |
-
|
| 395 |
-
|
| 396 |
-
|
| 397 |
-
|
| 398 |
-
|
| 399 |
-
|
| 400 |
-
|
| 401 |
-
|
| 402 |
-
|
| 403 |
-
|
| 404 |
-
|
| 405 |
-
|
| 406 |
-
|
| 407 |
-
|
| 408 |
-
|
| 409 |
-
|
| 410 |
-
|
| 411 |
-
|
| 412 |
-
|
| 413 |
-
|
| 414 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 415 |
if content:
|
|
|
|
| 416 |
full_response += content
|
| 417 |
|
| 418 |
-
#
|
| 419 |
-
if "<reasoning>" in
|
| 420 |
-
|
|
|
|
| 421 |
if len(parts) > 1:
|
| 422 |
-
|
| 423 |
-
|
| 424 |
-
|
| 425 |
-
|
| 426 |
-
|
| 427 |
-
|
| 428 |
-
|
| 429 |
-
|
| 430 |
-
|
| 431 |
-
if temp_move_str:
|
| 432 |
-
move_str = temp_move_str
|
| 433 |
-
current_reasoning_html += f'<p><strong>📍 Move chosen:</strong> Column {move_str.upper()}</p>'
|
| 434 |
|
| 435 |
-
|
|
|
|
|
|
|
| 436 |
|
| 437 |
-
#
|
| 438 |
-
|
| 439 |
-
|
| 440 |
-
|
| 441 |
-
|
| 442 |
-
|
| 443 |
-
|
| 444 |
-
# Final response with complete reasoning and move
|
| 445 |
-
if not reasoning_content:
|
| 446 |
-
try:
|
| 447 |
-
reasoning_content = full_response.split("<reasoning>")[1].split("</reasoning>")[0].strip()
|
| 448 |
-
except:
|
| 449 |
-
reasoning_content = "AI couldn't provide clear reasoning."
|
| 450 |
-
|
| 451 |
-
if not move_str:
|
| 452 |
-
move_str = extract_xml_move(full_response)
|
| 453 |
-
|
| 454 |
-
final_reasoning_html = f'''
|
| 455 |
-
<div id="ai-reasoning">
|
| 456 |
-
<div class="reasoning-box">
|
| 457 |
-
<p><strong>🤔 Reasoning:</strong></p>
|
| 458 |
-
<p>{reasoning_content}</p>
|
| 459 |
-
<p><strong>📍 Move chosen:</strong> Column {move_str.upper() if move_str else "?"}</p>
|
| 460 |
-
</div>
|
| 461 |
-
</div>
|
| 462 |
-
'''
|
| 463 |
-
|
| 464 |
-
# Make AI move
|
| 465 |
-
if move_str:
|
| 466 |
-
ai_col = game.parse_ai_move(move_str)
|
| 467 |
-
|
| 468 |
-
if ai_col != -1:
|
| 469 |
-
success, _ = game.make_move(ai_col)
|
| 470 |
-
if success:
|
| 471 |
-
# Check for AI winner
|
| 472 |
-
winner = game.check_winner()
|
| 473 |
-
if winner == 2:
|
| 474 |
-
game.game_over = True
|
| 475 |
yield [
|
| 476 |
render_board(game.board),
|
| 477 |
-
"
|
| 478 |
-
|
| 479 |
]
|
| 480 |
-
|
| 481 |
-
|
| 482 |
-
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 486 |
|
| 487 |
-
|
| 488 |
-
|
| 489 |
-
|
| 490 |
-
|
| 491 |
-
|
| 492 |
-
|
| 493 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 494 |
|
| 495 |
-
#
|
| 496 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 497 |
|
| 498 |
def reset_game():
|
| 499 |
game.board = np.zeros((6, 7))
|
|
@@ -509,13 +558,20 @@ def create_interface():
|
|
| 509 |
|
| 510 |
# Event handlers
|
| 511 |
for i, btn in enumerate(col_buttons):
|
|
|
|
| 512 |
btn.click(
|
| 513 |
-
fn=
|
| 514 |
inputs=[
|
| 515 |
gr.Number(value=i, visible=False),
|
| 516 |
temperature_slider
|
| 517 |
],
|
| 518 |
-
outputs=[board_display, status, reasoning_display]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 519 |
)
|
| 520 |
|
| 521 |
reset_btn.click(
|
|
|
|
| 4 |
from huggingface_hub import hf_hub_download
|
| 5 |
import numpy as np
|
| 6 |
from typing import List
|
| 7 |
+
import time
|
| 8 |
|
| 9 |
model = Llama(
|
| 10 |
model_path=hf_hub_download(
|
|
|
|
| 48 |
return match.group(1)
|
| 49 |
return ""
|
| 50 |
|
| 51 |
+
def extract_reasoning(text: str) -> str:
|
| 52 |
+
"""
|
| 53 |
+
Extract the reasoning part from the XML response.
|
| 54 |
+
"""
|
| 55 |
+
import re
|
| 56 |
+
match = re.search(r'<reasoning>(.*?)</reasoning>', text, re.DOTALL)
|
| 57 |
+
if match:
|
| 58 |
+
return match.group(1).strip()
|
| 59 |
+
return ""
|
| 60 |
+
|
| 61 |
def convert_moves_to_coordinate_list(moves_list: List[str]) -> str:
|
| 62 |
"""
|
| 63 |
Converts a list of moves to a coordinate list representation.
|
|
|
|
| 302 |
margin: 15px 0;
|
| 303 |
font-family: monospace;
|
| 304 |
min-height: 100px;
|
| 305 |
+
max-height: 400px;
|
| 306 |
+
overflow-y: auto;
|
| 307 |
+
white-space: pre-wrap;
|
| 308 |
}
|
| 309 |
.reasoning-box {
|
| 310 |
border-left: 4px solid #2196F3;
|
|
|
|
| 327 |
div.svelte-1nguped {
|
| 328 |
display: block;
|
| 329 |
}
|
| 330 |
+
.thinking-indicator {
|
| 331 |
+
color: #ffc107;
|
| 332 |
+
font-style: italic;
|
| 333 |
+
}
|
| 334 |
"""
|
| 335 |
|
| 336 |
with gr.Blocks(css=css) as interface:
|
|
|
|
| 370 |
info="Lower values make AI more deterministic, higher values more creative"
|
| 371 |
)
|
| 372 |
|
| 373 |
+
def handle_player_move(col, temperature=0.8):
|
| 374 |
if game.game_over:
|
| 375 |
return [
|
| 376 |
render_board(game.board),
|
|
|
|
| 404 |
game_state = game.format_game_state()
|
| 405 |
print(game_state)
|
| 406 |
|
| 407 |
+
# Prepare for AI's turn
|
| 408 |
+
reasoning_html = '''
|
| 409 |
+
<div id="ai-reasoning">
|
| 410 |
+
<div class="reasoning-box">
|
| 411 |
+
<p><strong>🤔 Reasoning:</strong></p>
|
| 412 |
+
<p class="thinking-indicator">AI is thinking...</p>
|
| 413 |
+
</div>
|
| 414 |
+
</div>
|
| 415 |
+
'''
|
| 416 |
+
|
| 417 |
+
return [
|
| 418 |
+
render_board(game.board),
|
| 419 |
+
"AI is thinking...",
|
| 420 |
+
reasoning_html
|
| 421 |
+
]
|
| 422 |
+
|
| 423 |
+
def process_ai_response(temperature=0.8):
|
| 424 |
+
if game.game_over or game.current_player != 2:
|
| 425 |
+
return [
|
| 426 |
+
render_board(game.board),
|
| 427 |
+
"Your turn!" if not game.game_over else "Game is over! Click New Game to play again.",
|
| 428 |
+
reasoning_display.value
|
| 429 |
+
]
|
| 430 |
+
|
| 431 |
+
game_state = game.format_game_state()
|
| 432 |
+
|
| 433 |
+
# Start streaming response
|
| 434 |
+
full_response = ""
|
| 435 |
+
reasoning_content = ""
|
| 436 |
+
move_str = ""
|
| 437 |
+
|
| 438 |
+
# Begin HTML
|
| 439 |
+
reasoning_html = '''
|
| 440 |
+
<div id="ai-reasoning">
|
| 441 |
+
<div class="reasoning-box">
|
| 442 |
+
<p><strong>🤔 Reasoning:</strong></p>
|
| 443 |
+
<p>
|
| 444 |
+
'''
|
| 445 |
+
|
| 446 |
+
# Create streaming session
|
| 447 |
+
response_stream = model.create_chat_completion(
|
| 448 |
+
messages=[
|
| 449 |
+
{"role": "system", "content": SYSTEM_PROMPT},
|
| 450 |
+
{"role": "user", "content": game_state}
|
| 451 |
+
],
|
| 452 |
+
temperature=temperature,
|
| 453 |
+
top_p=0.95,
|
| 454 |
+
max_tokens=1024,
|
| 455 |
+
stream=True
|
| 456 |
+
)
|
| 457 |
+
|
| 458 |
+
# Process stream
|
| 459 |
+
collected_text = ""
|
| 460 |
+
for chunk in response_stream:
|
| 461 |
+
if 'choices' in chunk and len(chunk['choices']) > 0:
|
| 462 |
+
delta = chunk['choices'][0]['delta']
|
| 463 |
+
if 'content' in delta:
|
| 464 |
+
content = delta['content']
|
| 465 |
if content:
|
| 466 |
+
collected_text += content
|
| 467 |
full_response += content
|
| 468 |
|
| 469 |
+
# Extract and update reasoning
|
| 470 |
+
if "<reasoning>" in collected_text and "</reasoning>" not in collected_text:
|
| 471 |
+
# Still collecting reasoning
|
| 472 |
+
parts = collected_text.split("<reasoning>")
|
| 473 |
if len(parts) > 1:
|
| 474 |
+
reasoning_content = parts[1]
|
| 475 |
+
current_html = reasoning_html + reasoning_content + '</p>'
|
| 476 |
+
if not move_str:
|
| 477 |
+
current_html += '</div></div>'
|
| 478 |
+
yield [
|
| 479 |
+
render_board(game.board),
|
| 480 |
+
"AI is thinking...",
|
| 481 |
+
current_html
|
| 482 |
+
]
|
|
|
|
|
|
|
|
|
|
| 483 |
|
| 484 |
+
# Check for complete reasoning
|
| 485 |
+
if "<reasoning>" in collected_text and "</reasoning>" in collected_text:
|
| 486 |
+
reasoning_content = extract_reasoning(collected_text)
|
| 487 |
|
| 488 |
+
# Check for move
|
| 489 |
+
new_move = extract_xml_move(collected_text)
|
| 490 |
+
if new_move and not move_str:
|
| 491 |
+
move_str = new_move
|
| 492 |
+
current_html = reasoning_html + reasoning_content + '</p>'
|
| 493 |
+
current_html += f'<p><strong>📍 Move chosen:</strong> Column {move_str.upper()}</p>'
|
| 494 |
+
current_html += '</div></div>'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 495 |
yield [
|
| 496 |
render_board(game.board),
|
| 497 |
+
"AI is making a move...",
|
| 498 |
+
current_html
|
| 499 |
]
|
| 500 |
+
|
| 501 |
+
# Final formatting
|
| 502 |
+
final_reasoning_html = f'''
|
| 503 |
+
<div id="ai-reasoning">
|
| 504 |
+
<div class="reasoning-box">
|
| 505 |
+
<p><strong>🤔 Reasoning:</strong></p>
|
| 506 |
+
<p>{reasoning_content if reasoning_content else "AI couldn't provide clear reasoning."}</p>
|
| 507 |
+
<p><strong>📍 Move chosen:</strong> Column {move_str.upper() if move_str else "?"}</p>
|
| 508 |
+
</div>
|
| 509 |
+
</div>
|
| 510 |
+
'''
|
| 511 |
+
|
| 512 |
+
# Make AI move
|
| 513 |
+
if move_str:
|
| 514 |
+
ai_col = game.parse_ai_move(move_str)
|
| 515 |
|
| 516 |
+
if ai_col != -1:
|
| 517 |
+
success, _ = game.make_move(ai_col)
|
| 518 |
+
if success:
|
| 519 |
+
# Check for AI winner
|
| 520 |
+
winner = game.check_winner()
|
| 521 |
+
if winner == 2:
|
| 522 |
+
game.game_over = True
|
| 523 |
+
yield [
|
| 524 |
+
render_board(game.board),
|
| 525 |
+
"🤖 AI wins! Better luck next time!",
|
| 526 |
+
final_reasoning_html
|
| 527 |
+
]
|
| 528 |
+
return
|
| 529 |
+
else:
|
| 530 |
+
# AI made a valid move but didn't win
|
| 531 |
+
game.current_player = 1
|
| 532 |
+
yield [
|
| 533 |
+
render_board(game.board),
|
| 534 |
+
"Your turn!",
|
| 535 |
+
final_reasoning_html
|
| 536 |
+
]
|
| 537 |
+
return
|
| 538 |
|
| 539 |
+
# If we get here, something went wrong with the AI move
|
| 540 |
+
game.game_over = True
|
| 541 |
+
yield [
|
| 542 |
+
render_board(game.board),
|
| 543 |
+
"AI made an invalid move! You win by default!",
|
| 544 |
+
'<div id="ai-reasoning">AI made an invalid move or provided an invalid response!</div>'
|
| 545 |
+
]
|
| 546 |
|
| 547 |
def reset_game():
|
| 548 |
game.board = np.zeros((6, 7))
|
|
|
|
| 558 |
|
| 559 |
# Event handlers
|
| 560 |
for i, btn in enumerate(col_buttons):
|
| 561 |
+
# First handle player move
|
| 562 |
btn.click(
|
| 563 |
+
fn=handle_player_move,
|
| 564 |
inputs=[
|
| 565 |
gr.Number(value=i, visible=False),
|
| 566 |
temperature_slider
|
| 567 |
],
|
| 568 |
+
outputs=[board_display, status, reasoning_display],
|
| 569 |
+
queue=False # Process player move immediately
|
| 570 |
+
).then( # Then process AI response with streaming
|
| 571 |
+
fn=process_ai_response,
|
| 572 |
+
inputs=[temperature_slider],
|
| 573 |
+
outputs=[board_display, status, reasoning_display],
|
| 574 |
+
streaming=True # Enable streaming for AI response
|
| 575 |
)
|
| 576 |
|
| 577 |
reset_btn.click(
|