File size: 13,767 Bytes
e9e7e23 |
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 |
#!/usr/bin/env python3
"""
Validation script for the quantized ONNX LazarusNLP IndoBERT model.
Checks model integrity, performance, and accuracy.
"""
import onnxruntime as ort
from transformers import AutoTokenizer
import numpy as np
import json
import os
import time
import sys
def check_files():
"""Check if all required files are present."""
print("π Checking required files...")
required_files = [
"model.onnx",
"tokenizer.json",
"tokenizer_config.json",
"special_tokens_map.json",
"vocab.txt",
"config.json",
"README.md"
]
missing_files = []
file_sizes = {}
for file in required_files:
if os.path.exists(file):
file_sizes[file] = os.path.getsize(file)
print(f"β
{file} ({file_sizes[file] / (1024*1024):.1f} MB)")
else:
missing_files.append(file)
print(f"β {file} - MISSING")
if missing_files:
print(f"\nβ Missing files: {missing_files}")
return False, {}
print("β
All required files present")
return True, file_sizes
def check_model_loading():
"""Test model and tokenizer loading."""
print("\nπ Testing model loading...")
try:
# Load tokenizer
start_time = time.time()
tokenizer = AutoTokenizer.from_pretrained("./")
tokenizer_time = time.time() - start_time
print(f"β
Tokenizer loaded ({tokenizer_time:.3f}s)")
# Load ONNX model
start_time = time.time()
session = ort.InferenceSession("model.onnx")
model_time = time.time() - start_time
print(f"β
ONNX model loaded ({model_time:.3f}s)")
# Check model inputs/outputs
inputs = session.get_inputs()
outputs = session.get_outputs()
print(f"β
Model inputs: {[inp.name for inp in inputs]}")
print(f"β
Model outputs: {[out.name for out in outputs]}")
return True, session, tokenizer
except Exception as e:
print(f"β Model loading failed: {e}")
return False, None, None
def test_basic_inference(session, tokenizer):
"""Test basic model inference."""
print("\nπ§ͺ Testing basic inference...")
test_texts = [
"Halo",
"Ini adalah tes sederhana.",
"Teknologi AI berkembang pesat di Indonesia.",
"Model machine learning membantu analisis data besar untuk memberikan insight yang berharga."
]
results = []
for i, text in enumerate(test_texts):
try:
# Tokenize
inputs = tokenizer(text, return_tensors="np", padding=True, truncation=True)
# Inference
start_time = time.time()
outputs = session.run(None, {
'input_ids': inputs['input_ids'],
'attention_mask': inputs['attention_mask']
})
inference_time = time.time() - start_time
# Check output
embeddings = outputs[0]
token_count = inputs['input_ids'].shape[1]
results.append({
'text': text,
'tokens': token_count,
'output_shape': embeddings.shape,
'inference_time': inference_time,
'has_nan': np.isnan(embeddings).any(),
'has_inf': np.isinf(embeddings).any(),
'output_range': [float(embeddings.min()), float(embeddings.max())]
})
print(f"β
Test {i+1}: {token_count} tokens β {embeddings.shape} ({inference_time:.4f}s)")
except Exception as e:
print(f"β Test {i+1} failed: {e}")
return False, []
return True, results
def test_batch_processing(session, tokenizer):
"""Test batch processing capability."""
print("\nπ¦ Testing batch processing...")
batch_texts = [
"Kalimat pertama untuk tes batch.",
"Ini adalah kalimat kedua yang sedikit lebih panjang.",
"Kalimat ketiga dengan panjang yang berbeda lagi untuk menguji padding.",
"Terakhir, kalimat keempat."
]
try:
# Batch processing
inputs = tokenizer(batch_texts, return_tensors="np", padding=True, truncation=True)
start_time = time.time()
outputs = session.run(None, {
'input_ids': inputs['input_ids'],
'attention_mask': inputs['attention_mask']
})
batch_time = time.time() - start_time
embeddings = outputs[0]
print(f"β
Batch shape: {embeddings.shape}")
print(f"β
Batch time: {batch_time:.4f}s")
print(f"β
Avg per item: {batch_time/len(batch_texts):.4f}s")
# Verify each item in batch
for i in range(len(batch_texts)):
item_embedding = embeddings[i]
if np.isnan(item_embedding).any() or np.isinf(item_embedding).any():
print(f"β Batch item {i} has invalid values")
return False
print("β
All batch items valid")
return True
except Exception as e:
print(f"β Batch processing failed: {e}")
return False
def test_edge_cases(session, tokenizer):
"""Test edge cases and error handling."""
print("\nπ§ Testing edge cases...")
edge_cases = [
("Empty string", ""),
("Single character", "a"),
("Numbers only", "123456789"),
("Punctuation", "!!!???..."),
("Mixed script", "Hello dunia 123 !@#"),
("Very long", "Kata " * 100), # ~400 characters
("Special tokens", "[CLS] [SEP] [MASK] [PAD] [UNK]")
]
passed = 0
total = len(edge_cases)
for name, text in edge_cases:
try:
inputs = tokenizer(text, return_tensors="np", padding=True, truncation=True)
outputs = session.run(None, {
'input_ids': inputs['input_ids'],
'attention_mask': inputs['attention_mask']
})
embeddings = outputs[0]
# Check for valid output
if embeddings.shape[0] == 1 and embeddings.shape[2] == 768:
if not (np.isnan(embeddings).any() or np.isinf(embeddings).any()):
print(f"β
{name}: {embeddings.shape}")
passed += 1
else:
print(f"β {name}: Invalid values (NaN/Inf)")
else:
print(f"β {name}: Wrong shape {embeddings.shape}")
except Exception as e:
print(f"β {name}: {e}")
print(f"\nβ
Edge cases passed: {passed}/{total}")
return passed == total
def performance_benchmark(session, tokenizer):
"""Run performance benchmark."""
print("\nβ‘ Performance benchmark...")
# Test different text lengths
test_cases = [
("Short (5 tokens)", "Halo dunia!"),
("Medium (15 tokens)", "Teknologi AI berkembang sangat pesat di era digital modern."),
("Long (50+ tokens)", " ".join(["Kalimat panjang dengan banyak kata untuk menguji performa model dalam memproses teks yang lebih kompleks dan detail."] * 2))
]
benchmark_results = {}
for name, text in test_cases:
times = []
token_count = len(tokenizer.encode(text))
# Warm up
inputs = tokenizer(text, return_tensors="np", padding=True, truncation=True)
session.run(None, {
'input_ids': inputs['input_ids'],
'attention_mask': inputs['attention_mask']
})
# Benchmark runs
for _ in range(20):
inputs = tokenizer(text, return_tensors="np", padding=True, truncation=True)
start_time = time.time()
outputs = session.run(None, {
'input_ids': inputs['input_ids'],
'attention_mask': inputs['attention_mask']
})
times.append(time.time() - start_time)
avg_time = np.mean(times)
std_time = np.std(times)
tokens_per_sec = token_count / avg_time
benchmark_results[name] = {
'avg_time': avg_time,
'std_time': std_time,
'token_count': token_count,
'tokens_per_sec': tokens_per_sec
}
print(f"β
{name}: {avg_time:.4f}s Β± {std_time:.4f}s ({tokens_per_sec:.1f} tokens/s)")
return benchmark_results
def check_config_consistency():
"""Check configuration file consistency."""
print("\nπ§ Checking configuration consistency...")
try:
# Load configurations
with open("config.json", "r") as f:
config = json.load(f)
with open("tokenizer_config.json", "r") as f:
tokenizer_config = json.load(f)
with open("export_config.json", "r") as f:
export_config = json.load(f)
# Check consistency
issues = []
# Max length consistency
model_max_pos = config.get("max_position_embeddings", 512)
tokenizer_max = tokenizer_config.get("model_max_length", 512)
if model_max_pos != tokenizer_max:
issues.append(f"Max length mismatch: model={model_max_pos}, tokenizer={tokenizer_max}")
# Check unlimited length setting
unlimited = export_config.get("unlimited_length", False)
dynamic_axes = export_config.get("dynamic_axes", False)
if unlimited and not dynamic_axes:
issues.append("Unlimited length enabled but dynamic_axes is False")
# Check quantization info
if "quantization" not in config:
issues.append("Missing quantization information in config")
if issues:
for issue in issues:
print(f"β οΈ {issue}")
else:
print("β
All configurations consistent")
return len(issues) == 0
except Exception as e:
print(f"β Config check failed: {e}")
return False
def generate_validation_report(results):
"""Generate validation report."""
print("\nπ VALIDATION REPORT")
print("=" * 60)
# Summary
all_passed = all([
results.get('files_ok', False),
results.get('loading_ok', False),
results.get('inference_ok', False),
results.get('batch_ok', False),
results.get('edge_cases_ok', False),
results.get('config_ok', False)
])
status = "β
PASSED" if all_passed else "β FAILED"
print(f"Overall Status: {status}")
print(f"\nFile Check: {'β
PASSED' if results.get('files_ok') else 'β FAILED'}")
print(f"Model Loading: {'β
PASSED' if results.get('loading_ok') else 'β FAILED'}")
print(f"Basic Inference: {'β
PASSED' if results.get('inference_ok') else 'β FAILED'}")
print(f"Batch Processing: {'β
PASSED' if results.get('batch_ok') else 'β FAILED'}")
print(f"Edge Cases: {'β
PASSED' if results.get('edge_cases_ok') else 'β FAILED'}")
print(f"Config Consistency: {'β
PASSED' if results.get('config_ok') else 'β FAILED'}")
# Performance summary
if 'benchmark' in results:
print(f"\nβ‘ PERFORMANCE SUMMARY")
for name, data in results['benchmark'].items():
print(f"{name}: {data['avg_time']:.4f}s ({data['tokens_per_sec']:.1f} tokens/s)")
# File sizes
if 'file_sizes' in results:
total_size = sum(results['file_sizes'].values()) / (1024*1024)
print(f"\nπ Total model size: {total_size:.1f} MB")
print("=" * 60)
return all_passed
def main():
"""Run complete model validation."""
print("π LazarusNLP IndoBERT ONNX - Model Validation")
print("=" * 60)
results = {}
# Check files
files_ok, file_sizes = check_files()
results['files_ok'] = files_ok
results['file_sizes'] = file_sizes
if not files_ok:
print("\nβ Validation failed: Missing required files")
return False
# Check model loading
loading_ok, session, tokenizer = check_model_loading()
results['loading_ok'] = loading_ok
if not loading_ok:
print("\nβ Validation failed: Model loading error")
return False
# Test inference
inference_ok, inference_results = test_basic_inference(session, tokenizer)
results['inference_ok'] = inference_ok
results['inference_results'] = inference_results
# Test batch processing
batch_ok = test_batch_processing(session, tokenizer)
results['batch_ok'] = batch_ok
# Test edge cases
edge_cases_ok = test_edge_cases(session, tokenizer)
results['edge_cases_ok'] = edge_cases_ok
# Performance benchmark
benchmark = performance_benchmark(session, tokenizer)
results['benchmark'] = benchmark
# Check config consistency
config_ok = check_config_consistency()
results['config_ok'] = config_ok
# Generate report
validation_passed = generate_validation_report(results)
# Save results
with open("validation_results.json", "w") as f:
json.dump(results, f, indent=2, default=str)
print(f"\nπΎ Validation results saved to validation_results.json")
if validation_passed:
print("π Model validation completed successfully!")
return True
else:
print("β Model validation failed!")
return False
if __name__ == "__main__":
success = main()
sys.exit(0 if success else 1) |