File size: 11,761 Bytes
1a7ee60 | 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 | """
Model Optimization Utilities
Convert PyTorch models to ONNX and apply quantization
"""
import torch
import onnx
import onnxruntime as ort
from transformers import AutoTokenizer
import numpy as np
import logging
import time
from typing import Dict, List
from pathlib import Path
from classifier_model import DocumentClassifier
from ner_model import DocumentNERModel
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class ModelOptimizer:
"""Optimize models for faster inference"""
@staticmethod
def convert_classifier_to_onnx(
pytorch_model_path: str,
output_path: str,
model_name: str = "microsoft/MiniLM-L6-H384-uncased",
num_labels: int = 4,
opset_version: int = 14
):
"""
Convert classifier to ONNX format
Args:
pytorch_model_path: Path to PyTorch weights
output_path: Path to save ONNX model
model_name: Base model name
num_labels: Number of classification labels
opset_version: ONNX opset version
"""
logger.info("Converting classifier to ONNX...")
# Load model
model = DocumentClassifier(model_name=model_name, num_labels=num_labels)
state_dict = torch.load(pytorch_model_path, map_location='cpu')
model.load_state_dict(state_dict)
model.eval()
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)
# Create dummy input
dummy_text = "This is a sample invoice for ONNX conversion"
dummy_input = tokenizer(
dummy_text,
max_length=512,
padding='max_length',
truncation=True,
return_tensors='pt'
)
# Export
torch.onnx.export(
model,
(dummy_input['input_ids'], dummy_input['attention_mask']),
output_path,
input_names=['input_ids', 'attention_mask'],
output_names=['logits'],
dynamic_axes={
'input_ids': {0: 'batch_size'},
'attention_mask': {0: 'batch_size'},
'logits': {0: 'batch_size'}
},
opset_version=opset_version,
do_constant_folding=True,
)
logger.info(f"Classifier exported to {output_path}")
# Verify
ModelOptimizer._verify_onnx_model(output_path)
@staticmethod
def convert_ner_to_onnx(
pytorch_model_path: str,
output_path: str,
model_name: str = "distilbert-base-uncased",
num_labels: int = 17,
opset_version: int = 14
):
"""
Convert NER model to ONNX format
Args:
pytorch_model_path: Path to PyTorch weights
output_path: Path to save ONNX model
model_name: Base model name
num_labels: Number of NER labels
opset_version: ONNX opset version
"""
logger.info("Converting NER model to ONNX...")
# Load model
model = DocumentNERModel(model_name=model_name, num_labels=num_labels)
state_dict = torch.load(pytorch_model_path, map_location='cpu')
model.load_state_dict(state_dict)
model.eval()
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)
# Create dummy input
dummy_text = "This is a sample invoice with number INV-12345"
dummy_input = tokenizer(
dummy_text,
max_length=512,
padding='max_length',
truncation=True,
return_tensors='pt'
)
# Export
torch.onnx.export(
model,
(dummy_input['input_ids'], dummy_input['attention_mask']),
output_path,
input_names=['input_ids', 'attention_mask'],
output_names=['logits'],
dynamic_axes={
'input_ids': {0: 'batch_size', 1: 'sequence_length'},
'attention_mask': {0: 'batch_size', 1: 'sequence_length'},
'logits': {0: 'batch_size', 1: 'sequence_length'}
},
opset_version=opset_version,
do_constant_folding=True,
)
logger.info(f"NER model exported to {output_path}")
# Verify
ModelOptimizer._verify_onnx_model(output_path)
@staticmethod
def _verify_onnx_model(onnx_path: str):
"""Verify ONNX model is valid"""
try:
onnx_model = onnx.load(onnx_path)
onnx.checker.check_model(onnx_model)
logger.info(f"ONNX model verified: {onnx_path}")
except Exception as e:
logger.error(f"ONNX verification failed: {str(e)}")
raise
@staticmethod
def quantize_onnx_model(
input_path: str,
output_path: str,
quantization_mode: str = "IntegerOps"
):
"""
Apply dynamic quantization to ONNX model
Reduces model size and improves CPU inference speed
Args:
input_path: Path to ONNX model
output_path: Path to save quantized model
quantization_mode: "IntegerOps" or "QLinearOps"
"""
from onnxruntime.quantization import quantize_dynamic, QuantType
logger.info(f"Quantizing ONNX model: {input_path}")
quantize_dynamic(
input_path,
output_path,
weight_type=QuantType.QInt8
)
logger.info(f"Quantized model saved to {output_path}")
# Compare sizes
original_size = Path(input_path).stat().st_size / (1024 * 1024)
quantized_size = Path(output_path).stat().st_size / (1024 * 1024)
logger.info(f"Original size: {original_size:.2f} MB")
logger.info(f"Quantized size: {quantized_size:.2f} MB")
logger.info(f"Size reduction: {(1 - quantized_size/original_size)*100:.1f}%")
class ONNXInferenceSession:
"""ONNX Runtime inference session wrapper"""
def __init__(self, model_path: str, providers: List[str] = None):
"""
Initialize ONNX Runtime session
Args:
model_path: Path to ONNX model
providers: Execution providers (e.g., ['CPUExecutionProvider'])
"""
if providers is None:
providers = ['CPUExecutionProvider']
self.session = ort.InferenceSession(model_path, providers=providers)
self.input_names = [inp.name for inp in self.session.get_inputs()]
self.output_names = [out.name for out in self.session.get_outputs()]
logger.info(f"ONNX session initialized: {model_path}")
logger.info(f"Inputs: {self.input_names}")
logger.info(f"Outputs: {self.output_names}")
def run(self, inputs: Dict[str, np.ndarray]) -> List[np.ndarray]:
"""
Run inference
Args:
inputs: Dictionary of input name -> numpy array
Returns:
List of output arrays
"""
# Prepare inputs
ort_inputs = {name: inputs[name] for name in self.input_names}
# Run
outputs = self.session.run(self.output_names, ort_inputs)
return outputs
def benchmark_models(
pytorch_model_path: str,
onnx_model_path: str,
model_type: str = "classifier",
num_runs: int = 100
):
"""
Benchmark PyTorch vs ONNX inference speed
Args:
pytorch_model_path: Path to PyTorch model
onnx_model_path: Path to ONNX model
model_type: "classifier" or "ner"
num_runs: Number of benchmark runs
"""
logger.info(f"Benchmarking {model_type} models...")
# Load tokenizer
if model_type == "classifier":
model_name = "microsoft/MiniLM-L6-H384-uncased"
num_labels = 4
ModelClass = DocumentClassifier
else:
model_name = "distilbert-base-uncased"
num_labels = 17
ModelClass = DocumentNERModel
tokenizer = AutoTokenizer.from_pretrained(model_name)
# Prepare sample input
sample_text = "This is a sample invoice with number INV-12345 dated 28/11/2025"
inputs = tokenizer(
sample_text,
max_length=512,
padding='max_length',
truncation=True,
return_tensors='pt'
)
# PyTorch model
pytorch_model = ModelClass(model_name=model_name, num_labels=num_labels)
pytorch_model.load_state_dict(torch.load(pytorch_model_path, map_location='cpu'))
pytorch_model.eval()
# ONNX model
onnx_session = ONNXInferenceSession(onnx_model_path)
# Warmup
for _ in range(10):
with torch.no_grad():
_ = pytorch_model(inputs['input_ids'], inputs['attention_mask'])
onnx_inputs = {
'input_ids': inputs['input_ids'].numpy(),
'attention_mask': inputs['attention_mask'].numpy()
}
_ = onnx_session.run(onnx_inputs)
# Benchmark PyTorch
pytorch_times = []
for _ in range(num_runs):
start = time.time()
with torch.no_grad():
_ = pytorch_model(inputs['input_ids'], inputs['attention_mask'])
pytorch_times.append(time.time() - start)
# Benchmark ONNX
onnx_times = []
for _ in range(num_runs):
start = time.time()
_ = onnx_session.run(onnx_inputs)
onnx_times.append(time.time() - start)
# Results
logger.info("\nBenchmark Results:")
logger.info(f"PyTorch - Mean: {np.mean(pytorch_times)*1000:.2f}ms, "
f"Std: {np.std(pytorch_times)*1000:.2f}ms")
logger.info(f"ONNX - Mean: {np.mean(onnx_times)*1000:.2f}ms, "
f"Std: {np.std(onnx_times)*1000:.2f}ms")
logger.info(f"Speedup: {np.mean(pytorch_times)/np.mean(onnx_times):.2f}x")
if __name__ == "__main__":
import sys
if len(sys.argv) < 2:
print("Usage:")
print(" Convert classifier: python model_optimizer.py convert_classifier <pytorch_model.pt> <output.onnx>")
print(" Convert NER: python model_optimizer.py convert_ner <pytorch_model.pt> <output.onnx>")
print(" Quantize: python model_optimizer.py quantize <input.onnx> <output_quantized.onnx>")
print(" Benchmark: python model_optimizer.py benchmark <pytorch_model.pt> <onnx_model.onnx> <classifier|ner>")
sys.exit(1)
command = sys.argv[1]
if command == "convert_classifier":
pytorch_path = sys.argv[2]
onnx_path = sys.argv[3]
ModelOptimizer.convert_classifier_to_onnx(pytorch_path, onnx_path)
elif command == "convert_ner":
pytorch_path = sys.argv[2]
onnx_path = sys.argv[3]
ModelOptimizer.convert_ner_to_onnx(pytorch_path, onnx_path)
elif command == "quantize":
input_path = sys.argv[2]
output_path = sys.argv[3]
ModelOptimizer.quantize_onnx_model(input_path, output_path)
elif command == "benchmark":
pytorch_path = sys.argv[2]
onnx_path = sys.argv[3]
model_type = sys.argv[4]
benchmark_models(pytorch_path, onnx_path, model_type)
else:
print(f"Unknown command: {command}")
|