Update validate_workflow_curated.py
Browse files- validate_workflow_curated.py +36 -20
validate_workflow_curated.py
CHANGED
|
@@ -39,6 +39,12 @@ TASK_INPUT_TYPES: dict[str, list[str]] = {
|
|
| 39 |
"translation": ["text"],
|
| 40 |
"text-classification": ["text"],
|
| 41 |
"question-answering": ["text"],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
"image-to-image": ["image"],
|
| 43 |
"image-to-text": ["image"],
|
| 44 |
"image-to-video": ["image"],
|
|
@@ -63,6 +69,12 @@ TASK_OUTPUT_TYPES: dict[str, list[str]] = {
|
|
| 63 |
"translation": ["text"],
|
| 64 |
"text-classification": ["json"],
|
| 65 |
"question-answering": ["text"],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
"image-to-image": ["image"],
|
| 67 |
"image-to-text": ["text"],
|
| 68 |
"image-to-video": ["video"],
|
|
@@ -168,7 +180,9 @@ def schema_cross_check(ep: dict, task: str) -> Optional[str]:
|
|
| 168 |
def default_smoke_inputs(task: str) -> list[Any]:
|
| 169 |
if task in ("text-to-image", "text-to-video", "text-to-3d", "text-to-speech", "text-to-audio"):
|
| 170 |
return ["a small red square"]
|
| 171 |
-
if task in ("text-generation", "summarization", "translation", "text-classification",
|
|
|
|
|
|
|
| 172 |
return ["hello world"]
|
| 173 |
if task in ("image-to-image", "image-to-text", "image-to-video", "image-to-3d",
|
| 174 |
"image-classification", "image-segmentation", "object-detection",
|
|
@@ -233,7 +247,6 @@ except Exception as e:
|
|
| 233 |
if proc.returncode != 0:
|
| 234 |
return False, proc.stderr.strip() or f"exit {proc.returncode}", latency
|
| 235 |
return True, None, latency
|
| 236 |
-
return True, None, int((time.monotonic() - started) * 1000)
|
| 237 |
|
| 238 |
|
| 239 |
WAKE_POLL_INTERVAL = 10 # seconds between runtime polls
|
|
@@ -366,17 +379,24 @@ def validate_model(entry: dict, hf_token: Optional[str]) -> dict:
|
|
| 366 |
def load_manifest(local_path: Optional[str], hf_token: Optional[str]) -> tuple[dict, str]:
|
| 367 |
if local_path:
|
| 368 |
with open(local_path, encoding="utf-8") as f:
|
| 369 |
-
|
| 370 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 371 |
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
)
|
| 378 |
-
with open(local, encoding="utf-8") as f:
|
| 379 |
-
return json.load(f), f"{CURATED_DATASET}/{CURATED_FILENAME}"
|
| 380 |
|
| 381 |
|
| 382 |
def upload_manifest(payload: dict, hf_token: str) -> None:
|
|
@@ -432,9 +452,9 @@ def main() -> int:
|
|
| 432 |
hf_token = os.environ.get("HF_TOKEN") or os.environ.get("HF_JOBS_TOKEN")
|
| 433 |
|
| 434 |
payload, src = load_manifest(args.source, hf_token)
|
| 435 |
-
items = payload.get("items")
|
| 436 |
if not isinstance(items, list):
|
| 437 |
-
logger.error("manifest at %s is malformed (no
|
| 438 |
return 2
|
| 439 |
if args.limit:
|
| 440 |
items = items[: args.limit]
|
|
@@ -476,18 +496,14 @@ def main() -> int:
|
|
| 476 |
order = {e.get("id", ""): i for i, e in enumerate(items)}
|
| 477 |
new_items.sort(key=lambda e: order.get(e.get("id", ""), len(items)))
|
| 478 |
|
| 479 |
-
out_payload = (
|
| 480 |
-
{**payload, "items": new_items, "fetched_at": now_iso()}
|
| 481 |
-
if isinstance(payload, dict)
|
| 482 |
-
else new_items
|
| 483 |
-
)
|
| 484 |
-
|
| 485 |
statuses: dict[str, int] = {}
|
| 486 |
for e in new_items:
|
| 487 |
s = (e.get("validation") or {}).get("status", "unknown")
|
| 488 |
statuses[s] = statuses.get(s, 0) + 1
|
| 489 |
logger.info("results: %s", statuses)
|
| 490 |
|
|
|
|
|
|
|
| 491 |
if args.dry_run:
|
| 492 |
json.dump(out_payload, sys.stdout, indent=2)
|
| 493 |
sys.stdout.write("\n")
|
|
|
|
| 39 |
"translation": ["text"],
|
| 40 |
"text-classification": ["text"],
|
| 41 |
"question-answering": ["text"],
|
| 42 |
+
"zero-shot-classification": ["text"],
|
| 43 |
+
"token-classification": ["text"],
|
| 44 |
+
"fill-mask": ["text"],
|
| 45 |
+
"visual-question-answering": ["image"],
|
| 46 |
+
"feature-extraction": ["text"],
|
| 47 |
+
"image-text-to-text": ["image"],
|
| 48 |
"image-to-image": ["image"],
|
| 49 |
"image-to-text": ["image"],
|
| 50 |
"image-to-video": ["image"],
|
|
|
|
| 69 |
"translation": ["text"],
|
| 70 |
"text-classification": ["json"],
|
| 71 |
"question-answering": ["text"],
|
| 72 |
+
"zero-shot-classification": ["json"],
|
| 73 |
+
"token-classification": ["json"],
|
| 74 |
+
"fill-mask": ["json"],
|
| 75 |
+
"visual-question-answering": ["text"],
|
| 76 |
+
"feature-extraction": ["json"],
|
| 77 |
+
"image-text-to-text": ["text"],
|
| 78 |
"image-to-image": ["image"],
|
| 79 |
"image-to-text": ["text"],
|
| 80 |
"image-to-video": ["video"],
|
|
|
|
| 180 |
def default_smoke_inputs(task: str) -> list[Any]:
|
| 181 |
if task in ("text-to-image", "text-to-video", "text-to-3d", "text-to-speech", "text-to-audio"):
|
| 182 |
return ["a small red square"]
|
| 183 |
+
if task in ("text-generation", "summarization", "translation", "text-classification",
|
| 184 |
+
"question-answering", "feature-extraction", "zero-shot-classification",
|
| 185 |
+
"token-classification", "fill-mask"):
|
| 186 |
return ["hello world"]
|
| 187 |
if task in ("image-to-image", "image-to-text", "image-to-video", "image-to-3d",
|
| 188 |
"image-classification", "image-segmentation", "object-detection",
|
|
|
|
| 247 |
if proc.returncode != 0:
|
| 248 |
return False, proc.stderr.strip() or f"exit {proc.returncode}", latency
|
| 249 |
return True, None, latency
|
|
|
|
| 250 |
|
| 251 |
|
| 252 |
WAKE_POLL_INTERVAL = 10 # seconds between runtime polls
|
|
|
|
| 379 |
def load_manifest(local_path: Optional[str], hf_token: Optional[str]) -> tuple[dict, str]:
|
| 380 |
if local_path:
|
| 381 |
with open(local_path, encoding="utf-8") as f:
|
| 382 |
+
data = json.load(f)
|
| 383 |
+
else:
|
| 384 |
+
from huggingface_hub import hf_hub_download
|
| 385 |
+
|
| 386 |
+
local = hf_hub_download(
|
| 387 |
+
repo_id=CURATED_DATASET,
|
| 388 |
+
filename=CURATED_FILENAME,
|
| 389 |
+
repo_type="dataset",
|
| 390 |
+
token=hf_token,
|
| 391 |
+
)
|
| 392 |
+
with open(local, encoding="utf-8") as f:
|
| 393 |
+
data = json.load(f)
|
| 394 |
|
| 395 |
+
# tolerate a bare-array manifest by wrapping it back up
|
| 396 |
+
if isinstance(data, list):
|
| 397 |
+
data = {"snapshot_version": 1, "items": data}
|
| 398 |
+
src = local_path or f"{CURATED_DATASET}/{CURATED_FILENAME}"
|
| 399 |
+
return data, src
|
|
|
|
|
|
|
|
|
|
| 400 |
|
| 401 |
|
| 402 |
def upload_manifest(payload: dict, hf_token: str) -> None:
|
|
|
|
| 452 |
hf_token = os.environ.get("HF_TOKEN") or os.environ.get("HF_JOBS_TOKEN")
|
| 453 |
|
| 454 |
payload, src = load_manifest(args.source, hf_token)
|
| 455 |
+
items = payload.get("items")
|
| 456 |
if not isinstance(items, list):
|
| 457 |
+
logger.error("manifest at %s is malformed (no items array)", src)
|
| 458 |
return 2
|
| 459 |
if args.limit:
|
| 460 |
items = items[: args.limit]
|
|
|
|
| 496 |
order = {e.get("id", ""): i for i, e in enumerate(items)}
|
| 497 |
new_items.sort(key=lambda e: order.get(e.get("id", ""), len(items)))
|
| 498 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 499 |
statuses: dict[str, int] = {}
|
| 500 |
for e in new_items:
|
| 501 |
s = (e.get("validation") or {}).get("status", "unknown")
|
| 502 |
statuses[s] = statuses.get(s, 0) + 1
|
| 503 |
logger.info("results: %s", statuses)
|
| 504 |
|
| 505 |
+
out_payload = {**payload, "items": new_items, "fetched_at": now_iso()}
|
| 506 |
+
|
| 507 |
if args.dry_run:
|
| 508 |
json.dump(out_payload, sys.stdout, indent=2)
|
| 509 |
sys.stdout.write("\n")
|