File size: 15,154 Bytes
e59ba0b |
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 |
import torch
from transformers import (
AutoModelForCausalLM,
AutoProcessor,
BitsAndBytesConfig,
GenerationConfig
)
from PIL import Image
import json
from typing import Optional, List, Dict, Any, Union
import time
from dataclasses import dataclass
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
@dataclass
class InferenceMetrics:
latency_ms: float
tokens_generated: int
tokens_per_second: float
memory_used_gb: float
input_tokens: int
total_tokens: int
class AdvancedHelionInference:
def __init__(
self,
model_name: str = "DeepXR/Helion-V2.0-Thinking",
quantization: Optional[str] = None,
device: str = "auto",
use_flash_attention: bool = True,
torch_compile: bool = False,
optimization_mode: str = "balanced"
):
logger.info(f"Initializing Helion-V2.0-Thinking with {optimization_mode} mode")
self.model_name = model_name
self.optimization_mode = optimization_mode
self.metrics_history = []
quantization_config = self._get_quantization_config(quantization)
logger.info("Loading processor...")
self.processor = AutoProcessor.from_pretrained(model_name)
logger.info("Loading model...")
self.model = AutoModelForCausalLM.from_pretrained(
model_name,
quantization_config=quantization_config,
device_map=device,
torch_dtype=torch.bfloat16 if quantization is None else None,
use_flash_attention_2=use_flash_attention,
trust_remote_code=True,
low_cpu_mem_usage=True
)
if torch_compile and quantization is None:
logger.info("Compiling model with torch.compile...")
self.model = torch.compile(self.model, mode="reduce-overhead")
self.model.eval()
self.generation_configs = {
"creative": GenerationConfig(
do_sample=True,
temperature=0.9,
top_p=0.95,
top_k=50,
repetition_penalty=1.15,
max_new_tokens=2048
),
"precise": GenerationConfig(
do_sample=True,
temperature=0.3,
top_p=0.85,
top_k=40,
repetition_penalty=1.05,
max_new_tokens=1024
),
"balanced": GenerationConfig(
do_sample=True,
temperature=0.7,
top_p=0.9,
top_k=50,
repetition_penalty=1.1,
max_new_tokens=1024
),
"code": GenerationConfig(
do_sample=True,
temperature=0.2,
top_p=0.9,
top_k=40,
repetition_penalty=1.05,
max_new_tokens=2048
)
}
logger.info("Model loaded successfully!")
def _get_quantization_config(self, quantization: Optional[str]) -> Optional[BitsAndBytesConfig]:
if quantization is None:
return None
quantization_configs = {
"4bit": BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type="nf4"
),
"8bit": BitsAndBytesConfig(
load_in_8bit=True
)
}
return quantization_configs.get(quantization)
def generate(
self,
prompt: str,
images: Optional[Union[Image.Image, List[Image.Image]]] = None,
mode: str = "balanced",
max_new_tokens: Optional[int] = None,
temperature: Optional[float] = None,
stream: bool = False,
return_metrics: bool = False,
**kwargs
) -> Union[str, tuple[str, InferenceMetrics]]:
if isinstance(images, Image.Image):
images = [images]
start_time = time.time()
initial_memory = torch.cuda.memory_allocated() / (1024**3) if torch.cuda.is_available() else 0
if images:
inputs = self.processor(
text=prompt,
images=images,
return_tensors="pt"
).to(self.model.device)
else:
inputs = self.processor(
text=prompt,
return_tensors="pt"
).to(self.model.device)
input_length = inputs['input_ids'].shape[1]
gen_config = self.generation_configs[mode].to_dict()
if max_new_tokens:
gen_config['max_new_tokens'] = max_new_tokens
if temperature:
gen_config['temperature'] = temperature
gen_config.update(kwargs)
with torch.no_grad():
if stream:
return self._generate_stream(inputs, gen_config, return_metrics)
else:
outputs = self.model.generate(
**inputs,
**gen_config,
pad_token_id=self.processor.tokenizer.eos_token_id
)
if torch.cuda.is_available():
torch.cuda.synchronize()
end_time = time.time()
latency = (end_time - start_time) * 1000
response = self.processor.decode(outputs[0], skip_special_tokens=True)
if response.startswith(prompt):
response = response[len(prompt):].strip()
tokens_generated = outputs.shape[1] - input_length
tokens_per_second = tokens_generated / ((end_time - start_time) if (end_time - start_time) > 0 else 1)
final_memory = torch.cuda.memory_allocated() / (1024**3) if torch.cuda.is_available() else 0
memory_used = final_memory - initial_memory
metrics = InferenceMetrics(
latency_ms=latency,
tokens_generated=tokens_generated,
tokens_per_second=tokens_per_second,
memory_used_gb=memory_used,
input_tokens=input_length,
total_tokens=outputs.shape[1]
)
self.metrics_history.append(metrics)
if return_metrics:
return response, metrics
return response
def _generate_stream(self, inputs, gen_config, return_metrics):
from transformers import TextIteratorStreamer
from threading import Thread
streamer = TextIteratorStreamer(
self.processor.tokenizer,
skip_special_tokens=True,
skip_prompt=True
)
gen_config['streamer'] = streamer
thread = Thread(
target=self.model.generate,
kwargs={**inputs, **gen_config}
)
thread.start()
for new_text in streamer:
yield new_text
thread.join()
def batch_generate(
self,
prompts: List[str],
images_list: Optional[List[Optional[Union[Image.Image, List[Image.Image]]]]] = None,
mode: str = "balanced",
**kwargs
) -> List[str]:
if images_list is None:
images_list = [None] * len(prompts)
all_inputs = []
for prompt, images in zip(prompts, images_list):
if images:
if isinstance(images, Image.Image):
images = [images]
inputs = self.processor(
text=prompt,
images=images,
return_tensors="pt",
padding=True
)
else:
inputs = self.processor(
text=prompt,
return_tensors="pt",
padding=True
)
all_inputs.append(inputs)
batch_inputs = {
k: torch.cat([inp[k] for inp in all_inputs], dim=0).to(self.model.device)
for k in all_inputs[0].keys()
}
gen_config = self.generation_configs[mode].to_dict()
gen_config.update(kwargs)
with torch.no_grad():
outputs = self.model.generate(
**batch_inputs,
**gen_config,
pad_token_id=self.processor.tokenizer.eos_token_id
)
responses = [
self.processor.decode(output, skip_special_tokens=True)
for output in outputs
]
return responses
def vision_qa(
self,
image: Image.Image,
question: str,
mode: str = "precise"
) -> str:
prompt = f"Question: {question}\nAnswer:"
return self.generate(prompt, images=image, mode=mode)
def analyze_image(
self,
image: Image.Image,
analysis_type: str = "detailed"
) -> str:
prompts = {
"detailed": "Provide a detailed description of this image, including objects, people, actions, setting, and any text visible.",
"quick": "Briefly describe what you see in this image.",
"technical": "Analyze this image from a technical perspective, including composition, lighting, colors, and quality.",
"ocr": "Extract all text visible in this image and organize it clearly."
}
prompt = prompts.get(analysis_type, prompts["detailed"])
return self.generate(prompt, images=image, mode="precise")
def code_generation(
self,
task: str,
language: str = "python",
include_tests: bool = False
) -> str:
prompt = f"Write {language} code for the following task:\n{task}"
if include_tests:
prompt += "\n\nInclude unit tests for the code."
return self.generate(prompt, mode="code", max_new_tokens=2048)
def function_call(
self,
user_query: str,
available_tools: List[Dict[str, Any]]
) -> Dict[str, Any]:
tools_str = json.dumps(available_tools, indent=2)
prompt = f"""Available tools:
{tools_str}
User query: {user_query}
Respond with a JSON object specifying which tool to use and with what parameters:
{{"tool": "tool_name", "parameters": {{"param": "value"}}}}
Response:"""
response = self.generate(prompt, mode="precise", temperature=0.2)
try:
import re
json_match = re.search(r'\{.*\}', response, re.DOTALL)
if json_match:
return json.loads(json_match.group())
return {"error": "No valid JSON found", "raw": response}
except json.JSONDecodeError as e:
return {"error": str(e), "raw": response}
def multi_modal_rag(
self,
query: str,
documents: List[str],
images: Optional[List[Image.Image]] = None
) -> str:
context = "\n\n".join([f"Document {i+1}:\n{doc}" for i, doc in enumerate(documents)])
prompt = f"""Context:\n{context}\n\nQuestion: {query}\n\nAnswer based on the provided context:"""
return self.generate(prompt, images=images, mode="precise", max_new_tokens=1024)
def get_metrics_summary(self) -> Dict[str, float]:
if not self.metrics_history:
return {}
return {
"avg_latency_ms": sum(m.latency_ms for m in self.metrics_history) / len(self.metrics_history),
"avg_tokens_per_second": sum(m.tokens_per_second for m in self.metrics_history) / len(self.metrics_history),
"avg_memory_used_gb": sum(m.memory_used_gb for m in self.metrics_history) / len(self.metrics_history),
"total_tokens_generated": sum(m.tokens_generated for m in self.metrics_history),
"num_requests": len(self.metrics_history)
}
def clear_cache(self):
if torch.cuda.is_available():
torch.cuda.empty_cache()
self.model.clear_cache() if hasattr(self.model, 'clear_cache') else None
logger.info("Cache cleared")
def main():
import argparse
parser = argparse.ArgumentParser(description="Advanced Helion-V2.0-Thinking Inference")
parser.add_argument("--model", type=str, default="DeepXR/Helion-V2.0-Thinking")
parser.add_argument("--quantization", type=str, choices=["4bit", "8bit", None], default=None)
parser.add_argument("--mode", type=str, default="balanced", choices=["creative", "precise", "balanced", "code"])
parser.add_argument("--prompt", type=str, help="Text prompt")
parser.add_argument("--image", type=str, help="Path to image file")
parser.add_argument("--stream", action="store_true", help="Enable streaming output")
parser.add_argument("--torch-compile", action="store_true", help="Use torch.compile")
parser.add_argument("--benchmark", action="store_true", help="Run benchmark")
args = parser.parse_args()
model = AdvancedHelionInference(
model_name=args.model,
quantization=args.quantization,
torch_compile=args.torch_compile
)
if args.benchmark:
print("Running benchmark...")
test_prompts = [
"Explain quantum computing in simple terms.",
"Write a Python function to calculate fibonacci numbers.",
"What are the main causes of climate change?"
]
for prompt in test_prompts:
response, metrics = model.generate(
prompt,
mode=args.mode,
return_metrics=True
)
print(f"\nPrompt: {prompt}")
print(f"Response: {response[:100]}...")
print(f"Metrics: {metrics}")
summary = model.get_metrics_summary()
print(f"\nBenchmark Summary:")
for key, value in summary.items():
print(f" {key}: {value:.2f}")
elif args.prompt:
image = Image.open(args.image) if args.image else None
if args.stream:
print("Streaming response:")
for text in model.generate(args.prompt, images=image, mode=args.mode, stream=True):
print(text, end="", flush=True)
print()
else:
response, metrics = model.generate(
args.prompt,
images=image,
mode=args.mode,
return_metrics=True
)
print(f"Response: {response}")
print(f"\nMetrics:")
print(f" Latency: {metrics.latency_ms:.2f}ms")
print(f" Tokens/sec: {metrics.tokens_per_second:.2f}")
print(f" Tokens generated: {metrics.tokens_generated}")
else:
print("Please provide --prompt or use --benchmark")
if __name__ == "__main__":
main() |