Spaces:
Running
Running
File size: 26,035 Bytes
27d0b7a 07d06f9 27d0b7a 3c45b5b 27d0b7a 07d06f9 27d0b7a 7a6f98c 37a6d2c 4787686 37a6d2c 27d0b7a 3c45b5b 27d0b7a 7a6f98c 27d0b7a b1cb912 4787686 27d0b7a 8d2763b 27d0b7a 3c45b5b 27d0b7a 07d06f9 27d0b7a 07d06f9 27d0b7a 07d06f9 27d0b7a 07d06f9 27d0b7a f0a58e4 27d0b7a f0a58e4 27d0b7a f0a58e4 27d0b7a f0a58e4 27d0b7a 7e9c2d9 27d0b7a 3c45b5b 27d0b7a 3c45b5b 27d0b7a 3c45b5b 27d0b7a 3c45b5b 27d0b7a 3c45b5b 27d0b7a 3c45b5b 27d0b7a 7e9c2d9 27d0b7a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 | """
MovieBench Preference Ranking User Study Application (Gradio Version)
A simplified Gradio web application for collecting human preference rankings
of AI-generated movies. For each story, presents results from different methods
side-by-side with shuffled anonymous labels, and collects preference ordering.
"""
import json
import logging
import os
import random
import threading
from datetime import datetime
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple
import gradio as gr
from huggingface_hub import CommitScheduler, snapshot_download
logging.getLogger("huggingface_hub._commit_scheduler").setLevel(logging.DEBUG)
logging.basicConfig(
format="%(asctime)s [%(name)s] %(levelname)s: %(message)s",
level=logging.INFO,
)
# ============================================================================
# Configuration
# ============================================================================
DATA_DIR = os.environ.get("DATA_DIR", "./data")
OUTPUT_DIR = os.environ.get("OUTPUT_DIR", "./results")
NUM_GROUPS = int(os.environ.get("NUM_GROUPS", "10"))
RESULTS_REPO_ID = os.environ.get("RESULTS_REPO_ID", "MovieBench/moviebench-results2")
DATA_REPO_ID = os.environ.get("DATA_REPO_ID", "MovieBench/moviebench-data2")
HF_TOKEN = os.environ.get("HF_TOKEN", None)
MAX_METHODS = 8
REFERENCE_AGENT = "EvoStoryGraph"
if DATA_REPO_ID and not Path(DATA_DIR).exists():
print(f"Downloading data from {DATA_REPO_ID} ...")
downloaded = snapshot_download(
repo_id=DATA_REPO_ID,
repo_type="dataset",
local_dir=DATA_DIR,
token=HF_TOKEN,
)
print(f"Data downloaded to {downloaded}")
Path(OUTPUT_DIR).mkdir(parents=True, exist_ok=True)
scheduler: Optional[CommitScheduler] = None
if RESULTS_REPO_ID:
print(f"Using scheduler for {RESULTS_REPO_ID} ...")
scheduler = CommitScheduler(
repo_id=RESULTS_REPO_ID,
repo_type="dataset",
folder_path=OUTPUT_DIR,
every=1,
path_in_repo="final_results/rk_pref",
token=HF_TOKEN,
)
# ============================================================================
# Data Loading Functions
# ============================================================================
def _load_story_scripts() -> Dict[str, str]:
"""Load original story scripts from vistory_test_lite.json (keyed by story id)."""
script_path = Path(DATA_DIR) / "vistory_test_lite.json"
if script_path.exists():
with open(script_path, "r", encoding="utf-8-sig") as f:
entries = json.load(f)
return {entry["id"]: entry["script"]["cn"] for entry in entries}
return {}
STORY_SCRIPTS = _load_story_scripts()
def load_summary() -> List[Dict[str, str]]:
"""Load summary.json that maps sample IDs to agents and story IDs."""
summary_path = Path(DATA_DIR) / "summary.json"
if summary_path.exists():
with open(summary_path, "r", encoding="utf-8-sig") as f:
return json.load(f)
return []
def get_available_samples() -> List[str]:
"""Get list of available sample directory IDs."""
data_path = Path(DATA_DIR)
if not data_path.exists():
return []
return sorted([d.name for d in data_path.iterdir() if d.is_dir()])
def get_stories_with_agents() -> Dict[str, List[Dict[str, str]]]:
"""Build mapping: story_id -> [{agent, shuffled_id}, ...]."""
summary = load_summary()
available = set(get_available_samples())
mapping: Dict[str, List[Dict[str, str]]] = {}
for entry in summary:
sid = entry["shuffled_id"]
if sid not in available:
continue
story_id = entry["story_id"]
mapping.setdefault(story_id, []).append({
"agent": entry["agent"],
"shuffled_id": sid,
})
for v in mapping.values():
v.sort(key=lambda x: x["agent"])
return mapping
def get_movie_video_path(shuffled_id: str) -> str:
"""Return the path to a sample's final movie video."""
p = Path(DATA_DIR) / shuffled_id / "final_video.mp4"
return str(p) if p.exists() else ""
def load_characters(sample_id: str) -> List[Dict]:
"""Load characters.json for a given sample."""
characters_path = Path(DATA_DIR) / sample_id / "characters.json"
if characters_path.exists():
with open(characters_path, "r", encoding="utf-8-sig") as f:
return json.load(f)
return []
def get_character_portraits(sample_id: str, characters: List[Dict]) -> List[Tuple[str, str]]:
"""Get character portrait paths. Returns list of (path, name) tuples.
Uses the convention: character_portraits/{idx}_{name}/front.png
"""
result = []
portraits_dir = Path(DATA_DIR).resolve() / sample_id / "character_portraits"
for char in characters:
idx = char.get("idx", "")
name = char.get("identifier_in_scene", "")
portrait_path = portraits_dir / f"{idx}_{name}" / "front.png"
if portrait_path.exists():
result.append((str(portrait_path), name))
return result
def get_reference_portraits(story_id: str) -> List[Tuple[str, str]]:
"""Get character portraits from the REFERENCE_AGENT sample for a story."""
stories_map = get_stories_with_agents()
entries = stories_map.get(story_id, [])
ref_sid = ""
for entry in entries:
if entry["agent"] == REFERENCE_AGENT:
ref_sid = entry["shuffled_id"]
break
if not ref_sid:
return []
characters = load_characters(ref_sid)
return get_character_portraits(ref_sid, characters)
_save_lock = threading.Lock()
# ============================================================================
# Group Management
# ============================================================================
def _partition_list(items: List, num_chunks: int) -> List[List]:
"""Split items into num_chunks chunks as evenly as possible."""
chunk_size, remainder = divmod(len(items), num_chunks)
chunks: List[List] = []
start = 0
for i in range(num_chunks):
end = start + chunk_size + (1 if i < remainder else 0)
chunks.append(items[start:end])
start = end
return chunks
def get_or_create_group_config(group_id: str) -> Dict[str, Any]:
"""Load existing group config or create a new one."""
group_dir = Path(OUTPUT_DIR) / f"group_{group_id}"
mapping_path = group_dir / "mapping.json"
if mapping_path.exists():
with open(mapping_path, "r", encoding="utf-8-sig") as f:
return json.load(f)
return create_group_config(group_id)
def create_group_config(group_id: str) -> Dict[str, Any]:
"""Create a group config with deterministic story partitioning and method shuffle.
Stories are shuffled with a fixed global seed and split into NUM_GROUPS
non-overlapping chunks. The agent display order is shuffled per-group
so that anonymous labels (Method A, B, ...) are consistent within a group
but differ across groups.
"""
group_dir = Path(OUTPUT_DIR) / f"group_{group_id}"
group_dir.mkdir(parents=True, exist_ok=True)
stories_map = get_stories_with_agents()
try:
group_index = (int(group_id) - 1) % NUM_GROUPS
except ValueError:
group_index = hash(group_id) % NUM_GROUPS
unique_stories = sorted(stories_map.keys())
story_rng = random.Random("moviebench_pref_story_partition")
story_rng.shuffle(unique_stories)
story_chunks = _partition_list(unique_stories, NUM_GROUPS)
selected_stories = story_chunks[group_index]
all_agents = set()
for story_id in selected_stories:
for entry in stories_map.get(story_id, []):
all_agents.add(entry["agent"])
all_agents_sorted = sorted(all_agents)
method_rng = random.Random(f"moviebench_pref_group_{group_id}")
shuffled_agents = list(all_agents_sorted)
method_rng.shuffle(shuffled_agents)
labels = [chr(ord("A") + i) for i in range(len(shuffled_agents))]
method_display_map = {}
for i, agent in enumerate(shuffled_agents):
method_display_map[f"Method {labels[i]}"] = agent
presentation_rng = random.Random(f"moviebench_pref_order_{group_id}")
story_order = list(selected_stories)
presentation_rng.shuffle(story_order)
config = {
"group_id": group_id,
"group_index": group_index,
"num_groups": NUM_GROUPS,
"created_at": datetime.now().isoformat(),
"stories": story_order,
"total_stories": len(unique_stories),
"stories_in_group": len(story_order),
"agents": all_agents_sorted,
"method_order": shuffled_agents,
"method_display_map": method_display_map,
}
with _save_lock:
with open(group_dir / "mapping.json", "w", encoding="utf-8") as f:
json.dump(config, f, indent=2, ensure_ascii=False)
return config
def _on_push_done(future):
"""Callback to surface push results/errors from the background thread."""
try:
result = future.result()
if result:
print(f"[CommitScheduler] Push succeeded: {result.commit_url}")
else:
print("[CommitScheduler] Push skipped: no changed files detected")
except Exception as e:
print(f"[CommitScheduler] Push FAILED: {e}")
def save_ranking_result(
group_id: str,
story_id: str,
evaluator_id: str,
method_display_map: Dict[str, str],
ranking: Dict[str, int],
comment: str,
) -> str:
"""Save a preference ranking result to JSON. Returns a status message."""
group_dir = Path(OUTPUT_DIR) / f"group_{group_id}"
story_dir = group_dir / story_id
filename = f"{story_id}_{evaluator_id}.json"
result_data = {
"evaluator_id": evaluator_id,
"group_id": group_id,
"timestamp": datetime.now().isoformat(),
"story_id": story_id,
"method_order": method_display_map,
"ranking": ranking,
"comment": comment,
}
filepath = story_dir / filename
with _save_lock:
story_dir.mkdir(parents=True, exist_ok=True)
with open(filepath, "w", encoding="utf-8") as f:
json.dump(result_data, f, indent=4, ensure_ascii=False)
print(f"[Save] Written {filepath}")
if scheduler is not None:
print("[CommitScheduler] Triggering immediate push after save ...")
future = scheduler.trigger()
future.add_done_callback(_on_push_done)
else:
print("[CommitScheduler] WARNING: scheduler is None — RESULTS_REPO_ID not set?")
return f"Saved to {filepath}"
# ============================================================================
# Gradio Interface
# ============================================================================
CUSTOM_CSS = """
.gradio-container {
max-width: 1600px !important;
margin-left: auto !important;
margin-right: auto !important;
}
.title-text {
text-align: center;
background: linear-gradient(135deg, #7c5cff 0%, #ff6b9d 100%);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
font-size: 2rem;
font-weight: 700;
margin-bottom: 1rem;
}
.method-label {
text-align: center;
font-size: 1.1rem;
font-weight: 600;
padding: 6px 0;
}
"""
def create_app():
"""Create the Gradio application."""
with gr.Blocks(
title="MovieBench: Preference Ranking",
css=CUSTOM_CSS,
theme=gr.themes.Soft(
primary_hue="purple",
secondary_hue="pink",
neutral_hue="slate",
),
) as app:
current_evaluator = gr.State("anonymous")
current_group = gr.State("")
group_config_state = gr.State({})
current_story_idx = gr.State(0)
gr.Markdown(
"# MovieBench: Preference Ranking",
elem_classes=["title-text"],
)
# ================================================================
# Tab 1: Setup
# ================================================================
with gr.Tab("Setup", id="tab_setup"):
gr.Markdown("### Enter your evaluator ID and group ID to begin")
with gr.Row():
evaluator_input = gr.Textbox(
label="Evaluator ID",
placeholder="Enter your name or ID",
value="anonymous",
scale=2,
)
group_input = gr.Textbox(
label="Group ID (auto-assigned, you may override)",
placeholder=f"Auto-assigned (1-{NUM_GROUPS})",
value="",
scale=2,
)
load_group_btn = gr.Button("Load / Create Group", variant="primary")
group_info = gr.Markdown("*Enter a Group ID and click 'Load / Create Group'*")
def load_group(group_id: str, evaluator_id: str):
if not group_id:
group_id = str(random.randint(1, NUM_GROUPS))
config = get_or_create_group_config(group_id)
stories = config.get("stories", [])
agents = config.get("agents", [])
method_map = config.get("method_display_map", {})
display_lines = ", ".join(sorted(method_map.keys()))
info_md = (
f"### Group `{group_id}` loaded "
f"(partition {config.get('group_index', 0) + 1}/{config.get('num_groups', NUM_GROUPS)})\n\n"
f"**Stories in group:** {len(stories)}/{config.get('total_stories', '?')}\n\n"
f"**Agents:** {len(agents)} ({', '.join(agents)})\n\n"
f"**Display labels:** {display_lines}\n\n"
f"**Story order:** {', '.join(stories)}\n\n"
f"**Created:** {config.get('created_at', 'N/A')}\n\n"
f"Go to the **Preference Evaluation** tab to start ranking."
)
return info_md, evaluator_id, group_id, config, gr.update(value=group_id)
load_group_btn.click(
load_group,
inputs=[group_input, evaluator_input],
outputs=[group_info, current_evaluator, current_group, group_config_state, group_input],
)
# ================================================================
# Tab 2: Preference Evaluation
# ================================================================
with gr.Tab("Preference Evaluation", id="tab_eval"):
gr.Markdown("### Rank the methods by preference for each story")
gr.Markdown(
"> **Note:** 不需要考虑音频质量、音画同步,重点关注**视觉一致性**、"
"**空间连贯性**、**叙事连贯性**、**剧本忠实度**、**视觉吸引力**。"
)
with gr.Row():
story_progress = gr.Markdown("**Progress:** Load a group first")
story_nav_prev = gr.Button("Previous Story", size="sm")
story_nav_next = gr.Button("Next Story", size="sm")
with gr.Accordion("Story Script", open=True):
story_script_display = gr.Markdown(
"*Load a group and go to this tab to see stories*"
)
with gr.Accordion("Character References", open=True):
char_gallery = gr.Gallery(
label="Characters (from EvoStoryGraph)",
columns=6,
height=180,
object_fit="contain",
)
gr.Markdown("---")
gr.Markdown("### Method Videos")
method_cols: List[gr.Column] = []
method_videos: List[gr.Video] = []
method_labels: List[gr.Markdown] = []
method_ranks: List[gr.Dropdown] = []
with gr.Row():
for i in range(MAX_METHODS):
with gr.Column(visible=False) as col:
lbl = gr.Markdown(
f"**Method {chr(ord('A') + i)}**",
elem_classes=["method-label"],
)
vid = gr.Video(
label=f"Method {chr(ord('A') + i)}",
height=300,
)
rank = gr.Dropdown(
label="Rank",
choices=[],
value=None,
interactive=True,
)
method_cols.append(col)
method_videos.append(vid)
method_labels.append(lbl)
method_ranks.append(rank)
gr.Markdown("---")
rank_comment = gr.Textbox(
label="Comment (optional)",
placeholder="Any additional notes about your ranking decision...",
lines=2,
)
with gr.Row():
submit_btn = gr.Button("Submit & Next Story", variant="primary")
eval_status = gr.Markdown("")
# ============================================================
# Helper functions
# ============================================================
def _build_story_display(story_idx: int, config: Dict[str, Any]):
"""Build all output values for displaying a given story.
Returns a flat list matching the outputs wired to the UI:
[progress_md, script_md, gallery_items,
col_0_visible, vid_0, lbl_0, rank_0_choices,
col_1_visible, vid_1, lbl_1, rank_1_choices,
... (MAX_METHODS times)]
"""
stories = config.get("stories", [])
method_order: List[str] = config.get("method_order", [])
method_display_map: Dict[str, str] = config.get("method_display_map", {})
stories_map = get_stories_with_agents()
num_methods = len(method_order)
rank_choices = [str(r) for r in range(1, num_methods + 1)]
if not stories or story_idx >= len(stories):
outputs: list = [
"**Progress:** No stories loaded",
"*Load a group first*",
[],
]
for _ in range(MAX_METHODS):
outputs.extend([
gr.update(visible=False),
None,
"",
gr.update(choices=[], value=None),
])
return outputs
story_id = stories[story_idx]
script_text = STORY_SCRIPTS.get(story_id, "(Script not available)")
progress_md = f"**Progress:** Story {story_idx + 1}/{len(stories)} (`{story_id}`)"
script_md = f"**Story ID:** `{story_id}`\n\n{script_text}"
gallery_items = get_reference_portraits(story_id)
agent_to_sid: Dict[str, str] = {}
for entry in stories_map.get(story_id, []):
agent_to_sid[entry["agent"]] = entry["shuffled_id"]
label_to_agent = {}
for label in sorted(method_display_map.keys()):
label_to_agent[label] = method_display_map[label]
sorted_labels = sorted(label_to_agent.keys())
outputs = [progress_md, script_md, gallery_items]
for i in range(MAX_METHODS):
if i < len(sorted_labels):
label = sorted_labels[i]
agent = label_to_agent[label]
sid = agent_to_sid.get(agent, "")
video_path = get_movie_video_path(sid) if sid else ""
outputs.extend([
gr.update(visible=True),
video_path if video_path else None,
f"**{label}**",
gr.update(choices=rank_choices, value=None),
])
else:
outputs.extend([
gr.update(visible=False),
None,
"",
gr.update(choices=[], value=None),
])
return outputs
def update_story_display(story_idx: int, config: Dict[str, Any]):
return _build_story_display(story_idx, config)
def go_prev_story(story_idx: int):
return max(0, story_idx - 1)
def go_next_story(story_idx: int, config: Dict[str, Any]):
stories = config.get("stories", [])
return min(len(stories) - 1, story_idx + 1) if stories else 0
def submit_ranking(
story_idx: int,
evaluator_id: str,
group_id: str,
config: Dict[str, Any],
comment: str,
*rank_values,
):
"""Validate and save the ranking, then advance to the next story."""
if not group_id or not config:
return "Please load a group first", story_idx, gr.update()
stories = config.get("stories", [])
if not stories or story_idx >= len(stories):
return "No stories available", story_idx, gr.update()
method_display_map = config.get("method_display_map", {})
sorted_labels = sorted(method_display_map.keys())
num_methods = len(sorted_labels)
ranking: Dict[str, int] = {}
used_ranks = set()
for i in range(num_methods):
val = rank_values[i] if i < len(rank_values) else None
if val is None or val == "":
return (
f"Please assign a rank to **{sorted_labels[i]}**",
story_idx,
gr.update(),
)
r = int(val)
if r in used_ranks:
return (
f"Duplicate rank {r} — each method must have a unique rank",
story_idx,
gr.update(),
)
used_ranks.add(r)
ranking[sorted_labels[i]] = r
story_id = stories[story_idx]
status = save_ranking_result(
group_id=group_id,
story_id=story_id,
evaluator_id=evaluator_id,
method_display_map=method_display_map,
ranking=ranking,
comment=comment or "",
)
next_idx = min(len(stories) - 1, story_idx + 1)
if next_idx == story_idx:
return (
f"{status}\n\nAll stories evaluated! Thank you!",
next_idx,
"",
)
return (
f"{status} | Moving to next story...",
next_idx,
"",
)
# ============================================================
# Wire up events
# ============================================================
display_outputs = [story_progress, story_script_display, char_gallery]
for i in range(MAX_METHODS):
display_outputs.extend([
method_cols[i],
method_videos[i],
method_labels[i],
method_ranks[i],
])
# When group config changes, reset to story 0
group_config_state.change(
lambda cfg: [0] + _build_story_display(0, cfg),
inputs=[group_config_state],
outputs=[current_story_idx] + display_outputs,
)
# When story idx changes, update display
current_story_idx.change(
update_story_display,
inputs=[current_story_idx, group_config_state],
outputs=display_outputs,
)
story_nav_prev.click(
go_prev_story,
inputs=[current_story_idx],
outputs=[current_story_idx],
)
story_nav_next.click(
go_next_story,
inputs=[current_story_idx, group_config_state],
outputs=[current_story_idx],
)
submit_inputs = [
current_story_idx,
current_evaluator,
current_group,
group_config_state,
rank_comment,
] + method_ranks
submit_btn.click(
submit_ranking,
inputs=submit_inputs,
outputs=[eval_status, current_story_idx, rank_comment],
)
def _assign_random_group():
return str(random.randint(1, NUM_GROUPS))
app.load(_assign_random_group, outputs=[group_input])
return app
# ============================================================================
# Main Entry Point
# ============================================================================
demo = create_app()
if __name__ == "__main__":
data_dir_abs = str(Path(DATA_DIR).resolve())
demo.launch(
server_name="0.0.0.0",
server_port=7860,
share=False,
show_error=True,
allowed_paths=[data_dir_abs],
)
|