Spaces:
Running
Running
File size: 11,474 Bytes
c47ec10 | 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 | from __future__ import annotations
import base64
import math
import re
from io import BytesIO
from typing import Any
from PIL import Image
DEFAULT_IMAGE_SIZE = (1024, 1024)
IMAGE_INPUT_TOKEN_MODEL = "gpt-5.4-mini"
PATCH_SIZE = 32
TILE_SIZE = 512
TILE_HIGH_SHORT_SIDE = 768
PATCH_1536_MODELS = (
"gpt-5.4-mini",
"gpt-5.4-nano",
"gpt-5-mini",
"gpt-5-nano",
"gpt-5.2",
"gpt-5.3-codex",
"gpt-5-codex-mini",
"gpt-5.1-codex-mini",
"gpt-5.2-codex",
"gpt-5.2-chat-latest",
"o4-mini",
"gpt-4.1-mini",
"gpt-4.1-nano",
)
PATCH_MULTIPLIERS = {
"gpt-5.4-mini": 1.62,
"gpt-5.4-nano": 2.46,
"gpt-5-mini": 1.62,
"gpt-5-nano": 2.46,
"gpt-4.1-mini": 1.62,
"gpt-4.1-nano": 2.46,
"o4-mini": 1.72,
}
def _model_name(model: str) -> str:
return str(model or "").strip().lower()
def image_size_from_bytes(data: bytes) -> tuple[int, int] | None:
if not data:
return None
try:
with Image.open(BytesIO(data)) as image:
width, height = image.size
except Exception:
return None
if width <= 0 or height <= 0:
return None
return int(width), int(height)
def _decode_data_url(value: str) -> bytes:
text = str(value or "").strip()
payload = text.split(",", 1)[1] if text.startswith("data:") and "," in text else text
return base64.b64decode(payload)
def image_size_from_data_url(value: str) -> tuple[int, int] | None:
try:
return image_size_from_bytes(_decode_data_url(value))
except Exception:
return None
def parse_image_size(size: object, default: tuple[int, int] = DEFAULT_IMAGE_SIZE) -> tuple[int, int]:
if isinstance(size, (tuple, list)) and len(size) >= 2:
try:
width = int(size[0])
height = int(size[1])
if width > 0 and height > 0:
return width, height
except (TypeError, ValueError):
pass
match = re.search(r"(\d{2,5})\D+(\d{2,5})", str(size or ""))
if not match:
return default
width, height = int(match.group(1)), int(match.group(2))
return (width, height) if width > 0 and height > 0 else default
def _patch_count(width: float, height: float) -> int:
return math.ceil(width / PATCH_SIZE) * math.ceil(height / PATCH_SIZE)
def _patch_multiplier(model: str) -> float:
name = _model_name(model)
for prefix, multiplier in PATCH_MULTIPLIERS.items():
if name.startswith(prefix):
return multiplier
return 1.0
def _patch_limits(model: str, detail: str) -> tuple[int, int] | None:
name = _model_name(model)
if any(name.startswith(prefix) for prefix in PATCH_1536_MODELS):
return 1536, 2048
if name.startswith("gpt-5.5"):
return (10000, 6000) if detail in {"auto", "original"} else (2500, 2048)
if name.startswith("gpt-5.4"):
return (10000, 6000) if detail == "original" else (2500, 2048)
return None
def _patch_tokens(width: int, height: int, model: str, detail: str) -> int:
multiplier = _patch_multiplier(model)
if detail == "low":
return math.ceil(256 * multiplier)
limits = _patch_limits(model, detail)
if limits is None:
return 0
patch_budget, max_dimension = limits
scale = min(1.0, max_dimension / max(width, height))
resized_width = width * scale
resized_height = height * scale
if _patch_count(resized_width, resized_height) > patch_budget:
shrink_factor = math.sqrt((PATCH_SIZE * PATCH_SIZE * patch_budget) / (resized_width * resized_height))
width_units = resized_width * shrink_factor / PATCH_SIZE
height_units = resized_height * shrink_factor / PATCH_SIZE
adjusted_shrink_factor = shrink_factor * min(
math.floor(width_units) / width_units if width_units else 1,
math.floor(height_units) / height_units if height_units else 1,
)
resized_width *= adjusted_shrink_factor
resized_height *= adjusted_shrink_factor
tokens = min(_patch_count(max(1, resized_width), max(1, resized_height)), patch_budget)
return math.ceil(tokens * multiplier)
def _tile_rates(model: str) -> tuple[int, int]:
name = _model_name(model)
if name in {"gpt-5", "gpt-5-chat-latest"}:
return 70, 140
if name.startswith("gpt-4o-mini"):
return 2833, 5667
if name.startswith(("o1", "o1-pro", "o3")):
return 75, 150
if name.startswith("computer-use-preview"):
return 65, 129
return 85, 170
def _tile_tokens(width: int, height: int, model: str, detail: str) -> int:
base_tokens, tile_tokens = _tile_rates(model)
if detail == "low":
return base_tokens
scale = min(1.0, 2048 / max(width, height))
resized_width = width * scale
resized_height = height * scale
short_side = min(resized_width, resized_height)
if short_side > 0:
scale = TILE_HIGH_SHORT_SIDE / short_side
resized_width *= scale
resized_height *= scale
tiles = math.ceil(resized_width / TILE_SIZE) * math.ceil(resized_height / TILE_SIZE)
return base_tokens + tiles * tile_tokens
def count_image_input_tokens(
width: int,
height: int,
model: str,
detail: str = "auto",
input_fidelity: str = "low",
) -> int:
if width <= 0 or height <= 0:
return 0
detail = str(detail or "auto").strip().lower() or "auto"
return _patch_tokens(width, height, IMAGE_INPUT_TOKEN_MODEL, detail)
def _part_size(part: dict[str, Any]) -> tuple[int, int] | None:
try:
width = int(part.get("width") or 0)
height = int(part.get("height") or 0)
except (TypeError, ValueError):
width = height = 0
if width > 0 and height > 0:
return width, height
data = part.get("data")
if isinstance(data, (bytes, bytearray)):
return image_size_from_bytes(bytes(data))
image_url = part.get("image_url")
if isinstance(image_url, dict):
image_url = image_url.get("url") or image_url.get("image_url")
if isinstance(image_url, str) and image_url.startswith("data:"):
return image_size_from_data_url(image_url)
source = part.get("source")
if isinstance(source, dict) and str(source.get("type") or "") == "base64":
try:
return image_size_from_bytes(base64.b64decode(str(source.get("data") or "")))
except Exception:
return None
return None
def count_image_content_tokens(content: object, model: str, default_detail: str = "auto") -> int:
if not isinstance(content, list):
return 0
total = 0
for part in content:
if not isinstance(part, dict):
continue
part_type = str(part.get("type") or "").strip()
if part_type not in {"image", "image_url", "input_image"} and not part.get("source"):
continue
size = _part_size(part)
if not size:
continue
total += count_image_input_tokens(
size[0],
size[1],
model,
str(part.get("detail") or default_detail or "auto"),
str(part.get("input_fidelity") or part.get("inputFidelity") or "low"),
)
return total
def count_image_inputs_tokens(images: object, model: str, default_detail: str = "auto") -> int:
if not images:
return 0
total = 0
entries = images if isinstance(images, list) else [images]
for image in entries:
data = image[0] if isinstance(image, tuple) and image else image
if not isinstance(data, (bytes, bytearray)):
continue
size = image_size_from_bytes(bytes(data))
if size:
total += count_image_input_tokens(size[0], size[1], model, default_detail)
return total
def count_generated_image_tokens(width: int, height: int, quality: str = "auto") -> int:
patches = _patch_count(width, height)
quality = str(quality or "auto").strip().lower()
if quality == "low":
return math.ceil(patches * 17 / 64)
if quality in {"high", "hd"}:
return math.ceil(patches * 65 / 16)
return math.ceil(patches * 33 / 32)
def count_image_output_tokens(size: object = None, quality: str = "auto", count: int = 1) -> int:
width, height = parse_image_size(size)
return max(0, int(count or 0)) * count_generated_image_tokens(width, height, quality)
def count_image_output_items_tokens(
items: object,
size: object = None,
quality: str = "auto",
) -> int:
if not isinstance(items, list) or not items:
return 0
fallback_size = parse_image_size(size)
total = 0
for item in items:
image_size = None
if isinstance(item, dict):
b64_json = str(item.get("b64_json") or "").strip()
if b64_json:
try:
image_size = image_size_from_bytes(base64.b64decode(b64_json))
except Exception:
image_size = None
width, height = image_size or fallback_size
total += count_generated_image_tokens(width, height, quality)
return total
def token_usage(
input_text_tokens: int = 0,
input_image_tokens: int = 0,
output_text_tokens: int = 0,
output_image_tokens: int = 0,
) -> dict[str, Any]:
input_tokens = max(0, int(input_text_tokens or 0)) + max(0, int(input_image_tokens or 0))
output_tokens = max(0, int(output_text_tokens or 0)) + max(0, int(output_image_tokens or 0))
return {
"input_tokens": input_tokens,
"output_tokens": output_tokens,
"total_tokens": input_tokens + output_tokens,
"input_tokens_details": {
"text_tokens": max(0, int(input_text_tokens or 0)),
"image_tokens": max(0, int(input_image_tokens or 0)),
"cached_tokens": 0,
},
"output_tokens_details": {
"text_tokens": max(0, int(output_text_tokens or 0)),
"image_tokens": max(0, int(output_image_tokens or 0)),
"reasoning_tokens": 0,
},
}
def image_usage(
input_text_tokens: int = 0,
input_image_tokens: int = 0,
output_tokens: int = 0,
) -> dict[str, Any]:
return token_usage(
input_text_tokens=input_text_tokens,
input_image_tokens=input_image_tokens,
output_image_tokens=output_tokens,
)
def chat_usage_from_image_usage(usage: dict[str, Any]) -> dict[str, Any]:
input_tokens = int(usage.get("input_tokens") or 0)
output_tokens = int(usage.get("output_tokens") or 0)
input_details = usage.get("input_tokens_details") if isinstance(usage.get("input_tokens_details"), dict) else {}
output_details = usage.get("output_tokens_details") if isinstance(usage.get("output_tokens_details"), dict) else {}
return {
"prompt_tokens": input_tokens,
"completion_tokens": output_tokens,
"total_tokens": input_tokens + output_tokens,
"prompt_tokens_details": {
"text_tokens": int(input_details.get("text_tokens") or 0),
"image_tokens": int(input_details.get("image_tokens") or 0),
"cached_tokens": int(input_details.get("cached_tokens") or 0),
},
"completion_tokens_details": {
"text_tokens": int(output_details.get("text_tokens") or 0),
"image_tokens": int(output_details.get("image_tokens") or 0),
"reasoning_tokens": int(output_details.get("reasoning_tokens") or 0),
},
}
|