Spaces:
Running
Running
File size: 15,212 Bytes
e2b8b61 | 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 | import hashlib
import json
import re
import time
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict, List, Optional, Sequence, Tuple, Union
from azure.identity import (
AzureCliCredential,
ChainedTokenCredential,
ManagedIdentityCredential,
get_bearer_token_provider,
)
from openai import AzureOpenAI
from tqdm import tqdm
from src.config import (
DEFAULT_MODEL,
MAX_TOKENS,
MAX_TOKENS_THINKING,
TEMPERATURE,
TRAPI_SCOPE,
build_azure_endpoint,
)
from src.model_registry import get_model_config
from src.generation_utils import fill_template_file
from src.generation_rates import estimate_cost
Message = Dict[str, str]
PromptMessages = List[Message]
class Generator:
def __init__(
self,
model_alias: Optional[str] = None,
use_cache: bool = True,
cache_path: Optional[Union[str, Path]] = None,
max_retries: int = 5,
retry_wait_seconds: int = 90,
) -> None:
self.model_alias = model_alias or DEFAULT_MODEL
self.model_config = get_model_config(self.model_alias)
self.use_cache = use_cache
self.max_retries = max_retries
self.retry_wait_seconds = retry_wait_seconds
# Usage tracking
self.usage_stats = {
"total_prompt_tokens": 0,
"total_completion_tokens": 0,
"total_cached_tokens": 0,
"total_requests": 0,
"cache_hits": 0,
}
# Request log file (append-only JSONL, one per model)
self.request_log_path = (
Path("generation_logs") / f"requests_{self.model_alias}.jsonl"
)
self.request_log_path.parent.mkdir(parents=True, exist_ok=True)
self.cache_path = (
Path(cache_path)
if cache_path
else Path("generation_logs/prompt_cache.json")
)
self.cache_path.parent.mkdir(parents=True, exist_ok=True)
self.client = self._build_client()
self.prompt_cache: Dict[str, str] = self._load_prompt_cache()
def _build_client(self) -> AzureOpenAI:
token_provider = get_bearer_token_provider(
ChainedTokenCredential(
AzureCliCredential(),
ManagedIdentityCredential(),
),
TRAPI_SCOPE,
)
endpoint = (
self.model_config.endpoint_override
if self.model_config.endpoint_override
else build_azure_endpoint()
)
return AzureOpenAI(
azure_endpoint=endpoint,
azure_ad_token_provider=token_provider,
api_version=self.model_config.api_version,
)
def _load_prompt_cache(self) -> Dict[str, str]:
if not self.use_cache:
return {}
if not self.cache_path.exists():
return {}
try:
return json.loads(self.cache_path.read_text(encoding="utf-8"))
except Exception:
return {}
def _save_prompt_cache(self) -> None:
if not self.use_cache:
return
self.cache_path.write_text(
json.dumps(self.prompt_cache, indent=2, ensure_ascii=False),
encoding="utf-8",
)
@staticmethod
def _make_cache_key(
model_alias: str,
messages: PromptMessages,
temperature: float,
max_completion_tokens: int,
response_format: Optional[Dict[str, Any]],
) -> str:
payload = {
"model_alias": model_alias,
"messages": messages,
"temperature": temperature,
"max_completion_tokens": max_completion_tokens,
"response_format": response_format,
}
raw = json.dumps(payload, sort_keys=True, ensure_ascii=False)
return hashlib.sha256(raw.encode("utf-8")).hexdigest()
def _chat_once(
self,
messages: PromptMessages,
temperature: float,
max_completion_tokens: int,
response_format: Optional[Dict[str, Any]],
) -> str:
kwargs: Dict[str, Any] = {
"model": self.model_config.deployment_name,
"messages": messages,
"max_completion_tokens": max_completion_tokens,
"temperature": temperature,
}
# Only pass structured response_format to models that support it.
if response_format is not None and self.model_config.is_openai_compatible:
kwargs["response_format"] = response_format
# Disable thinking for Qwen 3.5 models to avoid slow reasoning tokens
if "qwen" in self.model_alias.lower() and "3.5" in self.model_alias:
kwargs["extra_body"] = {"chat_template_kwargs": {"enable_thinking": False}}
response = self.client.chat.completions.create(**kwargs)
content = response.choices[0].message.content or ""
# Capture usage stats
usage = getattr(response, "usage", None)
if usage:
prompt_tokens = getattr(usage, "prompt_tokens", 0) or 0
completion_tokens = getattr(usage, "completion_tokens", 0) or 0
cached_tokens = 0
prompt_details = getattr(usage, "prompt_tokens_details", None)
if prompt_details:
cached_tokens = getattr(prompt_details, "cached_tokens", 0) or 0
self.usage_stats["total_prompt_tokens"] += prompt_tokens
self.usage_stats["total_completion_tokens"] += completion_tokens
self.usage_stats["total_cached_tokens"] += cached_tokens
self.usage_stats["total_requests"] += 1
self._log_request(
messages=messages,
temperature=temperature,
max_completion_tokens=max_completion_tokens,
response_format=response_format,
prompt_tokens=prompt_tokens,
completion_tokens=completion_tokens,
cached_tokens=cached_tokens,
)
return content
def _log_request(
self,
messages: PromptMessages,
temperature: float,
max_completion_tokens: int,
response_format: Optional[Dict[str, Any]],
prompt_tokens: int,
completion_tokens: int,
cached_tokens: int,
) -> None:
"""Append a single request record to the JSONL log."""
record = {
"timestamp": datetime.now(timezone.utc).isoformat(),
"model_alias": self.model_alias,
"deployment": self.model_config.deployment_name,
"temperature": temperature,
"max_completion_tokens": max_completion_tokens,
"has_response_format": response_format is not None,
"prompt_tokens": prompt_tokens,
"completion_tokens": completion_tokens,
"cached_tokens": cached_tokens,
"cost_usd": estimate_cost(
self.model_alias, prompt_tokens, completion_tokens, cached_tokens
),
}
with open(self.request_log_path, "a", encoding="utf-8") as f:
f.write(json.dumps(record, ensure_ascii=False) + "\n")
def chat(
self,
messages: PromptMessages,
temperature: Optional[float] = None,
max_completion_tokens: Optional[int] = None,
response_format: Optional[Dict[str, Any]] = None,
dont_use_cached: bool = False,
) -> str:
final_temperature = TEMPERATURE if temperature is None else temperature
if max_completion_tokens is not None:
final_max_tokens = max_completion_tokens
elif not self.model_config.is_openai_compatible:
final_max_tokens = MAX_TOKENS_THINKING
else:
final_max_tokens = MAX_TOKENS
cache_key = self._make_cache_key(
model_alias=self.model_alias,
messages=messages,
temperature=final_temperature,
max_completion_tokens=final_max_tokens,
response_format=response_format,
)
if self.use_cache and not dont_use_cached and cache_key in self.prompt_cache:
self.usage_stats["cache_hits"] += 1
return self.prompt_cache[cache_key]
last_error: Optional[Exception] = None
for attempt in range(1, self.max_retries + 1):
try:
content = self._chat_once(
messages=messages,
temperature=final_temperature,
max_completion_tokens=final_max_tokens,
response_format=response_format,
)
if self.use_cache:
self.prompt_cache[cache_key] = content
self._save_prompt_cache()
return content
except Exception as exc:
last_error = exc
is_last_attempt = attempt == self.max_retries
print(
f"[Generator] Model call failed on attempt "
f"{attempt}/{self.max_retries} for model '{self.model_alias}': {exc}"
)
if is_last_attempt:
break
print(
f"[Generator] Waiting {self.retry_wait_seconds} seconds before retry..."
)
time.sleep(self.retry_wait_seconds)
raise RuntimeError(
f"Model call failed after {self.max_retries} attempts "
f"for model '{self.model_alias}'."
) from last_error
def prompt(
self,
prompts: Sequence[PromptMessages],
temperature: Optional[float] = None,
max_completion_tokens: Optional[int] = None,
response_format: Optional[Dict[str, Any]] = None,
dont_use_cached: bool = False,
skip_failures: bool = False,
) -> List[Optional[str]]:
outputs: List[Optional[str]] = []
for prompt_messages in tqdm(
prompts,
desc=f"[Generator:{self.model_alias}]",
unit="req",
):
try:
output = self.chat(
messages=prompt_messages,
temperature=temperature,
max_completion_tokens=max_completion_tokens,
response_format=response_format,
dont_use_cached=dont_use_cached,
)
outputs.append(output)
except Exception as exc:
if skip_failures:
print(f"[Generator] Skipping failed item: {exc}")
outputs.append(None)
else:
raise
return outputs
def get_usage_summary(self) -> Dict[str, Any]:
"""Return usage stats with estimated cost."""
stats = dict(self.usage_stats)
stats["model_alias"] = self.model_alias
stats["estimated_cost_usd"] = estimate_cost(
self.model_alias,
stats["total_prompt_tokens"],
stats["total_completion_tokens"],
stats["total_cached_tokens"],
)
return stats
def print_usage_summary(self, stage: str = "") -> None:
"""Print a human-readable usage summary."""
s = self.get_usage_summary()
label = f"[{stage}] " if stage else ""
cost_str = (
f"${s['estimated_cost_usd']:.4f}"
if s["estimated_cost_usd"] is not None
else "N/A (no pricing)"
)
print(
f"{label}Usage for {s['model_alias']}: "
f"{s['total_prompt_tokens']} prompt tokens, "
f"{s['total_completion_tokens']} completion tokens, "
f"{s['total_cached_tokens']} cached tokens | "
f"{s['total_requests']} API calls, "
f"{s['cache_hits']} cache hits | "
f"Est. cost: {cost_str}"
)
def build_prompts(
self,
template_path: Union[str, Path],
data: Sequence[Dict[str, Any]],
) -> Tuple[List[PromptMessages], Optional[Dict[str, Any]]]:
prompts: List[PromptMessages] = []
shared_response_format: Optional[Dict[str, Any]] = None
for item in data:
messages, response_format = fill_template_file(str(template_path), item)
prompts.append(messages)
if shared_response_format is None:
shared_response_format = response_format
return prompts, shared_response_format
@staticmethod
def parse_json_response(text: str):
text = text.strip()
# Try normal parse first
try:
return json.loads(text)
except Exception:
pass
# Strip code fences (e.g. ```json ... ```) common in Gemma responses
fence_pattern = r"```(?:json)?\s*([\s\S]*?)```"
fence_match = re.search(fence_pattern, text)
if fence_match:
try:
return json.loads(fence_match.group(1).strip())
except Exception:
pass
# Strip common thinking/reasoning prefixes from models like Qwen
# Look for JSON after thinking blocks
thinking_patterns = [
r"(?s).*?</think>\s*",
r"(?s)^Thinking Process:.*?(?=\{)",
r"(?s)^<think>.*?</think>\s*",
]
cleaned = text
for pattern in thinking_patterns:
match = re.match(pattern, cleaned)
if match:
cleaned = cleaned[match.end():]
try:
return json.loads(cleaned.strip())
except Exception:
pass
# Fallback: find the last complete JSON object (most likely the actual output)
# Search from the end to avoid noise from thinking blocks
brace_positions = [i for i, c in enumerate(text) if c == '{']
for start in reversed(brace_positions):
depth = 0
in_string = False
escape_next = False
for i in range(start, len(text)):
ch = text[i]
if escape_next:
escape_next = False
continue
if ch == '\\' and in_string:
escape_next = True
continue
if ch == '"' and not escape_next:
in_string = not in_string
continue
if in_string:
continue
if ch == '{':
depth += 1
elif ch == '}':
depth -= 1
if depth == 0:
candidate = text[start:i + 1]
try:
return json.loads(candidate)
except Exception:
break
continue
# Original simple fallback
start = text.find("{")
end = text.rfind("}")
if start != -1 and end != -1 and end > start:
json_str = text[start:end + 1]
try:
return json.loads(json_str)
except Exception:
pass
raise ValueError(f"Model response was not valid JSON:\n{text[:500]}") |