Spaces:
Running
Running
File size: 27,288 Bytes
8819832 d9ffa28 8819832 c8175b6 2d83b8c c8175b6 0900fea 8819832 b697afb 8819832 d9ffa28 b697afb 2d83b8c 8819832 b697afb 018daa2 d9ffa28 c8175b6 d9ffa28 b697afb d9ffa28 b697afb d9ffa28 018daa2 d9ffa28 b697afb dc7f0d4 854b81e d9ffa28 854b81e d9ffa28 854b81e d9ffa28 854b81e d9ffa28 854b81e d9ffa28 854b81e d9ffa28 854b81e d9ffa28 854b81e dc7f0d4 9f91b88 d9ffa28 2d83b8c 8819832 2d83b8c 8819832 2d83b8c c8175b6 8819832 2d83b8c 8819832 2d83b8c 8819832 018daa2 8819832 854b81e 8819832 2d83b8c 854b81e 2d83b8c 1eec007 b697afb 8819832 2d83b8c 8819832 2d83b8c 8819832 d9ffa28 2d83b8c d9ffa28 2d83b8c b697afb 2d83b8c 475172d 2d83b8c 018daa2 d9ffa28 018daa2 d9ffa28 b697afb 854b81e d9ffa28 018daa2 2d83b8c 8819832 854b81e 1eec007 854b81e 2d83b8c 854b81e 8819832 854b81e 8819832 854b81e 8819832 2d83b8c 854b81e 8819832 854b81e 8819832 854b81e 2d83b8c 854b81e 8819832 854b81e 2d83b8c 854b81e 8819832 2d83b8c 8819832 2d83b8c 8819832 2d83b8c 8819832 2d83b8c 8819832 854b81e 0c757a9 854b81e 0c757a9 854b81e 0c757a9 8819832 854b81e 8819832 854b81e 0c757a9 8819832 854b81e b697afb 854b81e b697afb 854b81e d05fd64 854b81e b697afb 854b81e b697afb 854b81e d05fd64 854b81e b697afb 854b81e b697afb 854b81e b697afb 854b81e b697afb 854b81e b697afb 854b81e c8175b6 |
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 |
import json
import uuid
import datetime
from typing import Dict, Optional, Any, List
from huggingface_hub import HfApi, ModelCard
from urllib.parse import urlparse
from .utils import calculate_completeness_score
class AIBOMGenerator:
def __init__(
self,
hf_token: Optional[str] = None,
inference_model_url: Optional[str] = None,
use_inference: bool = True,
cache_dir: Optional[str] = None,
use_best_practices: bool = True, # Added parameter for industry-neutral scoring
):
self.hf_api = HfApi(token=hf_token)
self.inference_model_url = inference_model_url
self.use_inference = use_inference
self.cache_dir = cache_dir
self.enhancement_report = None # Store enhancement report as instance variable
self.use_best_practices = use_best_practices # Store best practices flag
def generate_aibom(
self,
model_id: str,
output_file: Optional[str] = None,
include_inference: Optional[bool] = None,
use_best_practices: Optional[bool] = None, # Added parameter for industry-neutral scoring
) -> Dict[str, Any]:
try:
model_id = self._normalise_model_id(model_id)
use_inference = include_inference if include_inference is not None else self.use_inference
# Use method parameter if provided, otherwise use instance variable
use_best_practices = use_best_practices if use_best_practices is not None else self.use_best_practices
model_info = self._fetch_model_info(model_id)
model_card = self._fetch_model_card(model_id)
# Store original metadata before any AI enhancement
original_metadata = self._extract_structured_metadata(model_id, model_info, model_card)
# Create initial AIBOM with original metadata
original_aibom = self._create_aibom_structure(model_id, original_metadata)
# Calculate initial score with industry-neutral approach if enabled
original_score = calculate_completeness_score(original_aibom, validate=True, use_best_practices=use_best_practices)
# Final metadata starts with original metadata
final_metadata = original_metadata.copy() if original_metadata else {}
# Apply AI enhancement if requested
ai_enhanced = False
ai_model_name = None
if use_inference and self.inference_model_url:
try:
# Extract additional metadata using AI
enhanced_metadata = self._extract_unstructured_metadata(model_card, model_id)
# If we got enhanced metadata, merge it with original
if enhanced_metadata:
ai_enhanced = True
ai_model_name = "BERT-base-uncased" # Will be replaced with actual model name
# Merge enhanced metadata with original (enhanced takes precedence)
for key, value in enhanced_metadata.items():
if value is not None and (key not in final_metadata or not final_metadata[key]):
final_metadata[key] = value
except Exception as e:
print(f"Error during AI enhancement: {e}")
# Continue with original metadata if enhancement fails
# Create final AIBOM with potentially enhanced metadata
aibom = self._create_aibom_structure(model_id, final_metadata)
# Calculate final score with industry-neutral approach if enabled
final_score = calculate_completeness_score(aibom, validate=True, use_best_practices=use_best_practices)
# Ensure metadata.properties exists
if "metadata" in aibom and "properties" not in aibom["metadata"]:
aibom["metadata"]["properties"] = []
# Note: Quality score information is no longer added to the AIBOM metadata
# This was removed as requested by the user
if output_file:
with open(output_file, 'w') as f:
json.dump(aibom, f, indent=2)
# Create enhancement report for UI display and store as instance variable
self.enhancement_report = {
"ai_enhanced": ai_enhanced,
"ai_model": ai_model_name if ai_enhanced else None,
"original_score": original_score,
"final_score": final_score,
"improvement": round(final_score["total_score"] - original_score["total_score"], 2) if ai_enhanced else 0
}
# Return only the AIBOM to maintain compatibility with existing code
return aibom
except Exception as e:
print(f"Error generating AIBOM: {e}")
# Return a minimal valid AIBOM structure in case of error
return self._create_minimal_aibom(model_id)
def _create_minimal_aibom(self, model_id: str) -> Dict[str, Any]:
"""Create a minimal valid AIBOM structure in case of errors"""
return {
"bomFormat": "CycloneDX",
"specVersion": "1.6",
"serialNumber": f"urn:uuid:{str(uuid.uuid4())}",
"version": 1,
"metadata": {
"timestamp": datetime.datetime.utcnow().isoformat() + "Z",
"tools": {
"components": [{
"bom-ref": "pkg:generic/@cybeats/[email protected]",
"type": "application",
"name": "aetheris-aibom-generator",
"version": "0.1.0",
"manufacturer": {
"name": "Aetheris AI"
}
}]
},
"component": {
"bom-ref": f"pkg:generic/{model_id.replace('/', '%2F')}@1.0",
"type": "application",
"name": model_id.split("/")[-1],
"description": f"AI model {model_id}",
"version": "1.0",
"purl": f"pkg:generic/{model_id.replace('/', '%2F')}@1.0",
"copyright": "NOASSERTION"
}
},
"components": [{
"bom-ref": f"pkg:huggingface/{model_id.replace('/', '/')}@1.0",
"type": "machine-learning-model",
"name": model_id.split("/")[-1],
"version": "1.0",
"purl": f"pkg:huggingface/{model_id.replace('/', '/')}@1.0"
}],
"dependencies": [{
"ref": f"pkg:generic/{model_id.replace('/', '%2F')}@1.0",
"dependsOn": [f"pkg:huggingface/{model_id.replace('/', '/')}@1.0"]
}]
}
def get_enhancement_report(self):
"""Return the enhancement report from the last generate_aibom call"""
return self.enhancement_report
def _fetch_model_info(self, model_id: str) -> Dict[str, Any]:
try:
return self.hf_api.model_info(model_id)
except Exception as e:
print(f"Error fetching model info for {model_id}: {e}")
return {}
# ---- new helper ---------------------------------------------------------
@staticmethod
def _normalise_model_id(raw_id: str) -> str:
"""
Accept either 'owner/model' or a full URL like
'https://huggingface.co/owner/model'. Return 'owner/model'.
"""
if raw_id.startswith(("http://", "https://")):
path = urlparse(raw_id).path.lstrip("/")
# path can contain extra segments (e.g. /commit/...), keep first two
parts = path.split("/")
if len(parts) >= 2:
return "/".join(parts[:2])
return path
return raw_id
# -------------------------------------------------------------------------
def _fetch_model_card(self, model_id: str) -> Optional[ModelCard]:
try:
return ModelCard.load(model_id)
except Exception as e:
print(f"Error fetching model card for {model_id}: {e}")
return None
def _create_aibom_structure(
self,
model_id: str,
metadata: Dict[str, Any],
) -> Dict[str, Any]:
# Extract owner and model name from model_id
parts = model_id.split("/")
group = parts[0] if len(parts) > 1 else ""
name = parts[1] if len(parts) > 1 else parts[0]
# Get version from metadata or use default
version = metadata.get("commit", "1.0")
aibom = {
"bomFormat": "CycloneDX",
"specVersion": "1.6",
"serialNumber": f"urn:uuid:{str(uuid.uuid4())}",
"version": 1,
"metadata": self._create_metadata_section(model_id, metadata),
"components": [self._create_component_section(model_id, metadata)],
"dependencies": [
{
"ref": f"pkg:generic/{model_id.replace('/', '%2F')}@{version}",
"dependsOn": [f"pkg:huggingface/{model_id.replace('/', '/')}@{version}"]
}
]
}
# Add downloadLocation if available
if metadata and "commit_url" in metadata:
# Add external reference for downloadLocation
if "externalReferences" not in aibom:
aibom["externalReferences"] = []
aibom["externalReferences"].append({
"type": "distribution",
"url": f"https://huggingface.co/{model_id}"
})
return aibom
def _extract_structured_metadata(
self,
model_id: str,
model_info: Dict[str, Any],
model_card: Optional[ModelCard],
) -> Dict[str, Any]:
metadata = {}
if model_info:
try:
metadata.update({
"name": model_info.modelId.split("/")[-1] if hasattr(model_info, "modelId") else model_id.split("/")[-1],
"author": model_info.author if hasattr(model_info, "author") else None,
"tags": model_info.tags if hasattr(model_info, "tags") else [],
"pipeline_tag": model_info.pipeline_tag if hasattr(model_info, "pipeline_tag") else None,
"downloads": model_info.downloads if hasattr(model_info, "downloads") else 0,
"last_modified": model_info.lastModified if hasattr(model_info, "lastModified") else None,
"commit": model_info.sha[:7] if hasattr(model_info, "sha") and model_info.sha else None,
"commit_url": f"https://huggingface.co/{model_id}/commit/{model_info.sha}" if hasattr(model_info, "sha") and model_info.sha else None,
})
except Exception as e:
print(f"Error extracting model info metadata: {e}")
if model_card and hasattr(model_card, "data") and model_card.data:
try:
card_data = model_card.data.to_dict() if hasattr(model_card.data, "to_dict") else {}
metadata.update({
"language": card_data.get("language"),
"license": card_data.get("license"),
"library_name": card_data.get("library_name"),
"base_model": card_data.get("base_model"),
"datasets": card_data.get("datasets"),
"model_name": card_data.get("model_name"),
"tags": card_data.get("tags", metadata.get("tags", [])),
"description": card_data.get("model_summary", None)
})
if hasattr(model_card.data, "eval_results") and model_card.data.eval_results:
metadata["eval_results"] = model_card.data.eval_results
except Exception as e:
print(f"Error extracting model card metadata: {e}")
metadata["ai:type"] = "Transformer"
metadata["ai:task"] = metadata.get("pipeline_tag", "Text Generation")
metadata["ai:framework"] = "PyTorch" if "transformers" in metadata.get("library_name", "") else "Unknown"
# Add fields for industry-neutral scoring (silently aligned with SPDX)
metadata["primaryPurpose"] = metadata.get("ai:task", "Text Generation")
metadata["suppliedBy"] = metadata.get("author", "Unknown")
# Add typeOfModel field
metadata["typeOfModel"] = metadata.get("ai:type", "Transformer")
return {k: v for k, v in metadata.items() if v is not None}
def _extract_unstructured_metadata(self, model_card: Optional[ModelCard], model_id: str) -> Dict[str, Any]:
"""
Extract additional metadata from model card using BERT model.
This is a placeholder implementation that would be replaced with actual BERT inference.
In a real implementation, this would:
1. Extract text from model card
2. Use BERT to identify key information
3. Structure the extracted information
For now, we'll simulate this with some basic extraction logic.
"""
enhanced_metadata = {}
# In a real implementation, we would use a BERT model here
# Since we can't install the required libraries due to space constraints,
# we'll simulate the enhancement with a placeholder implementation
if model_card and hasattr(model_card, "text") and model_card.text:
try:
card_text = model_card.text
# Simulate BERT extraction with basic text analysis
# In reality, this would be done with NLP models
# Extract description if missing
if card_text and "description" not in enhanced_metadata:
# Take first paragraph that's longer than 20 chars as description
paragraphs = [p.strip() for p in card_text.split('\n\n')]
for p in paragraphs:
if len(p) > 20 and not p.startswith('#'):
enhanced_metadata["description"] = p
break
# Extract limitations if present
if "limitations" not in enhanced_metadata:
if "## Limitations" in card_text:
limitations_section = card_text.split("## Limitations")[1].split("##")[0].strip()
if limitations_section:
enhanced_metadata["limitations"] = limitations_section
# Extract ethical considerations if present
if "ethical_considerations" not in enhanced_metadata:
for heading in ["## Ethical Considerations", "## Ethics", "## Bias"]:
if heading in card_text:
section = card_text.split(heading)[1].split("##")[0].strip()
if section:
enhanced_metadata["ethical_considerations"] = section
break
# Extract risks if present
if "risks" not in enhanced_metadata:
if "## Risks" in card_text:
risks_section = card_text.split("## Risks")[1].split("##")[0].strip()
if risks_section:
enhanced_metadata["risks"] = risks_section
# Extract datasets if present
if "datasets" not in enhanced_metadata:
datasets = []
if "## Dataset" in card_text or "## Datasets" in card_text:
dataset_section = ""
if "## Dataset" in card_text:
dataset_section = card_text.split("## Dataset")[1].split("##")[0].strip()
elif "## Datasets" in card_text:
dataset_section = card_text.split("## Datasets")[1].split("##")[0].strip()
if dataset_section:
# Simple parsing to extract dataset names
lines = dataset_section.split("\n")
for line in lines:
if line.strip() and not line.startswith("#"):
datasets.append({
"type": "dataset",
"name": line.strip().split()[0] if line.strip().split() else "Unknown",
"description": line.strip()
})
if datasets:
enhanced_metadata["datasets"] = datasets
except Exception as e:
print(f"Error extracting unstructured metadata: {e}")
return enhanced_metadata
def _create_metadata_section(self, model_id: str, metadata: Dict[str, Any]) -> Dict[str, Any]:
timestamp = datetime.datetime.utcnow().isoformat() + "Z"
# Get version from metadata or use default
version = metadata.get("commit", "1.0")
# Create tools section with components array
tools = {
"components": [{
"bom-ref": "pkg:generic/@cybeats/[email protected]",
"type": "application",
"name": "aetheris-aibom-generator",
"version": "1.0",
"manufacturer": {
"name": "Aetheris AI"
}
}]
}
# Create authors array
authors = []
if "author" in metadata and metadata["author"]:
authors.append({
"name": metadata["author"]
})
# Create component section for metadata
component = {
"bom-ref": f"pkg:generic/{model_id.replace('/', '%2F')}@{version}",
"type": "application",
"name": metadata.get("name", model_id.split("/")[-1]),
"description": metadata.get("description", f"AI model {model_id}"),
"version": version,
"purl": f"pkg:generic/{model_id.replace('/', '%2F')}@{version}"
}
# Add authors to component if available
if authors:
component["authors"] = authors
# Add publisher and supplier if author is available
if "author" in metadata and metadata["author"]:
component["publisher"] = metadata["author"]
component["supplier"] = {
"name": metadata["author"]
}
component["manufacturer"] = {
"name": metadata["author"]
}
# Add copyright
component["copyright"] = "NOASSERTION"
# Create properties array for additional metadata
properties = []
for key, value in metadata.items():
if key not in ["name", "author", "license", "description", "commit"] and value is not None:
if isinstance(value, (list, dict)):
if not isinstance(value, str):
value = json.dumps(value)
properties.append({"name": key, "value": str(value)})
# Assemble metadata section
metadata_section = {
"timestamp": timestamp,
"tools": tools,
"component": component
}
if properties:
metadata_section["properties"] = properties
return metadata_section
def _create_component_section(self, model_id: str, metadata: Dict[str, Any]) -> Dict[str, Any]:
# Extract owner and model name from model_id
parts = model_id.split("/")
group = parts[0] if len(parts) > 1 else ""
name = parts[1] if len(parts) > 1 else parts[0]
# Get version from metadata or use default
version = metadata.get("commit", "1.0")
# Create PURL with version information if commit is available
purl = f"pkg:huggingface/{model_id.replace('/', '/')}"
if "commit" in metadata:
purl = f"{purl}@{metadata['commit']}"
else:
purl = f"{purl}@{version}"
component = {
"bom-ref": f"pkg:huggingface/{model_id.replace('/', '/')}@{version}",
"type": "machine-learning-model",
"group": group,
"name": name,
"version": version,
"purl": purl
}
# Add licenses if available
if "license" in metadata:
component["licenses"] = [{
"license": {
"id": metadata["license"],
"url": self._get_license_url(metadata["license"])
}
}]
# Add description if available
if "description" in metadata:
component["description"] = metadata["description"]
# Add external references
external_refs = [{
"type": "website",
"url": f"https://huggingface.co/{model_id}"
}]
if "commit_url" in metadata:
external_refs.append({
"type": "vcs",
"url": metadata["commit_url"]
})
component["externalReferences"] = external_refs
# Add authors, publisher, supplier, manufacturer
if "author" in metadata and metadata["author"]:
component["authors"] = [{"name": metadata["author"]}]
component["publisher"] = metadata["author"]
component["supplier"] = {
"name": metadata["author"],
"url": [f"https://huggingface.co/{metadata['author']}"]
}
component["manufacturer"] = {
"name": metadata["author"],
"url": [f"https://huggingface.co/{metadata['author']}"]
}
# Add copyright
component["copyright"] = "NOASSERTION"
# Add model card section
component["modelCard"] = self._create_model_card_section(metadata)
return component
def _create_model_card_section(self, metadata: Dict[str, Any]) -> Dict[str, Any]:
model_card_section = {}
# Add quantitative analysis section
if "eval_results" in metadata:
model_card_section["quantitativeAnalysis"] = {
"performanceMetrics": metadata["eval_results"],
"graphics": {} # Empty graphics object as in the example
}
else:
model_card_section["quantitativeAnalysis"] = {"graphics": {}}
# Add properties section
properties = []
for key, value in metadata.items():
if key in ["author", "library_name", "license", "downloads", "likes", "tags", "created_at", "last_modified"]:
properties.append({"name": key, "value": str(value)})
if properties:
model_card_section["properties"] = properties
# Create model parameters section
model_parameters = {}
# Add outputs array
model_parameters["outputs"] = [{"format": "generated-text"}]
# Add task
model_parameters["task"] = metadata.get("pipeline_tag", "text-generation")
# Add architecture information
model_parameters["architectureFamily"] = "llama" if "llama" in metadata.get("name", "").lower() else "transformer"
model_parameters["modelArchitecture"] = f"{metadata.get('name', 'Unknown')}ForCausalLM"
# Add datasets array with proper structure
if "datasets" in metadata:
datasets = []
if isinstance(metadata["datasets"], list):
for dataset in metadata["datasets"]:
if isinstance(dataset, str):
datasets.append({
"type": "dataset",
"name": dataset,
"description": f"Dataset used for training {metadata.get('name', 'the model')}"
})
elif isinstance(dataset, dict) and "name" in dataset:
# Ensure dataset has the required structure
dataset_entry = {
"type": dataset.get("type", "dataset"),
"name": dataset["name"],
"description": dataset.get("description", f"Dataset: {dataset['name']}")
}
datasets.append(dataset_entry)
elif isinstance(metadata["datasets"], str):
datasets.append({
"type": "dataset",
"name": metadata["datasets"],
"description": f"Dataset used for training {metadata.get('name', 'the model')}"
})
if datasets:
model_parameters["datasets"] = datasets
# Add inputs array
model_parameters["inputs"] = [{"format": "text"}]
# Add model parameters to model card section
model_card_section["modelParameters"] = model_parameters
# Add considerations section
considerations = {}
for k in ["limitations", "ethical_considerations", "bias", "risks"]:
if k in metadata:
considerations[k] = metadata[k]
if considerations:
model_card_section["considerations"] = considerations
return model_card_section
def _get_license_url(self, license_id: str) -> str:
"""Get the URL for a license based on its SPDX ID."""
license_urls = {
"Apache-2.0": "https://www.apache.org/licenses/LICENSE-2.0",
"MIT": "https://opensource.org/licenses/MIT",
"BSD-3-Clause": "https://opensource.org/licenses/BSD-3-Clause",
"GPL-3.0": "https://www.gnu.org/licenses/gpl-3.0.en.html",
"CC-BY-4.0": "https://creativecommons.org/licenses/by/4.0/",
"CC-BY-SA-4.0": "https://creativecommons.org/licenses/by-sa/4.0/",
"CC-BY-NC-4.0": "https://creativecommons.org/licenses/by-nc/4.0/",
"CC-BY-ND-4.0": "https://creativecommons.org/licenses/by-nd/4.0/",
"CC-BY-NC-SA-4.0": "https://creativecommons.org/licenses/by-nc-sa/4.0/",
"CC-BY-NC-ND-4.0": "https://creativecommons.org/licenses/by-nc-nd/4.0/",
"LGPL-3.0": "https://www.gnu.org/licenses/lgpl-3.0.en.html",
"MPL-2.0": "https://www.mozilla.org/en-US/MPL/2.0/",
}
return license_urls.get(license_id, "https://spdx.org/licenses/")
|