File size: 11,476 Bytes
d6a54fa |
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 |
import os
import subprocess
import sys
import argparse
from pathlib import Path
from concurrent.futures import (
ProcessPoolExecutor,
as_completed,
)
from zipnn_compress_file import compress_file
sys.path.append(
os.path.abspath(
os.path.join(
os.path.dirname(__file__), ".."
)
)
)
KB = 1024
MB = 1024 * 1024
GB = 1024 * 1024 * 1024
RED = "\033[91m"
YELLOW = "\033[93m"
GREEN = "\033[92m"
RESET = "\033[0m"
def check_and_install_zipnn():
try:
import zipnn
except ImportError:
print("zipnn not found. Installing...")
subprocess.check_call(
[
sys.executable,
"-m",
"pip",
"install",
"zipnn",
"--upgrade",
]
)
import zipnn
def parse_streaming_chunk_size(
streaming_chunk_size,
):
if str(streaming_chunk_size).isdigit():
final = int(streaming_chunk_size)
else:
size_value = int(
streaming_chunk_size[:-2]
)
size_unit = streaming_chunk_size[
-2
].lower()
if size_unit == "k":
final = KB * size_value
elif size_unit == "m":
final = MB * size_value
elif size_unit == "g":
final = GB * size_value
else:
raise ValueError(
f"Invalid size unit: {size_unit}. Use 'k', 'm', or 'g'."
)
return final
def replace_in_file(file_path, old: str, new: str) -> None:
"""Given a file_path, replace all occurrences of `old` with `new` inpalce."""
with open(file_path, 'r') as file:
file_data = file.read()
file_data = file_data.replace(old, new)
with open(file_path, 'w') as file:
file.write(file_data)
def compress_files_with_suffix(
suffix,
dtype="",
streaming_chunk_size=1048576,
path=".",
delete=False,
r=False,
force=False,
max_processes=1,
hf_cache=False,
model="",
branch="main",
):
import zipnn
overwrite_first=True
file_list = []
streaming_chunk_size = (
parse_streaming_chunk_size(
streaming_chunk_size
)
)
if model:
if not hf_cache:
raise ValueError(
"Must specify --hf_cache when using --model"
)
try:
from huggingface_hub import scan_cache_dir
except ImportError:
raise ImportError(
"huggingface_hub not found. Please pip install huggingface_hub."
)
cache = scan_cache_dir()
repo = next((repo for repo in cache.repos if repo.repo_id == model), None)
if repo is not None:
print(f"Found repo {model} in cache")
# Get the latest revision path
hash = ''
try:
with open(os.path.join(repo.repo_path, 'refs', branch), "r") as ref:
hash = ref.read()
except FileNotFoundError:
raise FileNotFoundError(f"Branch {branch} not found in repo {model}")
path = os.path.join(repo.repo_path, 'snapshots', hash)
directories_to_search = (
os.walk(path)
if r
else [(path, [], os.listdir(path))]
)
files_found = False
for root, _, files in directories_to_search:
for file_name in files:
if file_name.endswith(suffix):
compressed_path = (
file_name + ".znn"
)
if not force and os.path.exists(
compressed_path
):
#
if overwrite_first:
overwrite_first=False
user_input = (
input(
f"Compressed files already exists; Would you like to overwrite them all (y/n)? "
)
.strip()
.lower()
)
if user_input not in (
"y",
"yes",
):
print(
f"No forced overwriting."
)
else:
print(
f"Overwriting all compressed files."
)
force=True
#
if not force:
user_input = (
input(
f"{compressed_path} already exists; overwrite (y/n)? "
)
.strip()
.lower()
)
if user_input not in (
"y",
"yes",
):
print(
f"Skipping {file_name}..."
)
continue
files_found = True
full_path = os.path.join(
root, file_name
)
file_list.append(full_path)
if file_list and hf_cache:
try:
from transformers.utils import (
SAFE_WEIGHTS_INDEX_NAME,
WEIGHTS_INDEX_NAME
)
except ImportError:
raise ImportError(
"Transformers not found. Please pip install transformers."
)
if os.path.exists(os.path.join(path, SAFE_WEIGHTS_INDEX_NAME)):
print(f"{YELLOW}Fixing Hugging Face model json...{RESET}")
blob_name = os.path.join(path, os.readlink(os.path.join(path, SAFE_WEIGHTS_INDEX_NAME)))
replace_in_file(
file_path=blob_name,
old=f"{suffix}",
new=f"{suffix}.znn"
)
elif os.path.exists(os.path.join(path, WEIGHTS_INDEX_NAME)):
print(f"{YELLOW}Fixing Hugging Face model json...{RESET}")
blob_name = os.path.join(path, os.readlink(os.path.join(path, WEIGHTS_INDEX_NAME)))
replace_in_file(
file_path=blob_name,
old=f"{suffix}",
new=f"{suffix}.znn"
)
with ProcessPoolExecutor(
max_workers=max_processes
) as executor:
future_to_file = {
executor.submit(
compress_file,
file,
dtype,
streaming_chunk_size,
delete,
True,
hf_cache,
): file
for file in file_list[:max_processes]
}
file_list = file_list[max_processes:]
while future_to_file:
for future in as_completed(
future_to_file
):
file = future_to_file.pop(future)
try:
future.result()
except Exception as exc:
print(
f"{RED}File {file} generated an exception: {exc}{RESET}"
)
if file_list:
next_file = file_list.pop(0)
future_to_file[
executor.submit(
compress_file,
next_file,
dtype,
streaming_chunk_size,
delete,
True,
hf_cache,
)
] = next_file
if not files_found:
print(
f"{RED}No files with the suffix '{suffix}' found.{RESET}"
)
print(f"{GREEN}All files compressed{RESET}")
if __name__ == "__main__":
if len(sys.argv) < 2:
print(
"Usage: python compress_files.py <suffix>"
)
print(
"Example: python compress_files.py 'safetensors'"
)
sys.exit(1)
parser = argparse.ArgumentParser(
description="Enter a suffix to compress, (optional) dtype, (optional) streaming chunk size, (optional) path to files."
)
parser.add_argument(
"suffix",
type=str,
help="Specify the file suffix to compress all files with that suffix. If a single file name is provided, only that file will be compressed.",
)
parser.add_argument(
"--float32",
action="store_true",
help="A flag that triggers float32 compression",
)
parser.add_argument(
"--streaming_chunk_size",
type=str,
help="An optional streaming chunk size. The format is int (for size in Bytes) or int+KB/MB/GB. Default is 1MB",
)
parser.add_argument(
"--path",
type=str,
help="Path to files to compress",
)
parser.add_argument(
"--delete",
action="store_true",
help="A flag that triggers deletion of a single file instead of compression",
)
parser.add_argument(
"-r",
action="store_true",
help="A flag that triggers recursive search on all subdirectories",
)
parser.add_argument(
"--recursive",
action="store_true",
help="A flag that triggers recursive search on all subdirectories",
)
parser.add_argument(
"--force",
action="store_true",
help="A flag that forces overwriting when compressing.",
)
parser.add_argument(
"--max_processes",
type=int,
help="The amount of maximum processes.",
)
parser.add_argument(
"--hf_cache",
action="store_true",
help="A flag that indicates if the file is in the Hugging Face cache. Must either specify --model or --path to the model's snapshot cache.",
)
parser.add_argument(
"--model",
type=str,
help="Only when using --hf_cache, specify the model name or path. E.g. 'ibm-granite/granite-7b-instruct'",
)
parser.add_argument(
"--model_branch",
type=str,
default="main",
help="Only when using --model, specify the model branch. Default is 'main'",
)
args = parser.parse_args()
optional_kwargs = {}
if args.float32:
optional_kwargs["dtype"] = 32
if args.streaming_chunk_size is not None:
optional_kwargs[
"streaming_chunk_size"
] = args.streaming_chunk_size
if args.path is not None:
optional_kwargs["path"] = args.path
if args.delete:
optional_kwargs["delete"] = args.delete
if args.r or args.recursive:
optional_kwargs["r"] = args.r
if args.force:
optional_kwargs["force"] = args.force
if args.max_processes:
optional_kwargs["max_processes"] = (
args.max_processes
)
if args.hf_cache:
optional_kwargs["hf_cache"] = args.hf_cache
if args.model:
optional_kwargs["model"] = args.model
if args.model_branch:
optional_kwargs[
"branch"
] = args.model_branch
check_and_install_zipnn()
compress_files_with_suffix(
args.suffix, **optional_kwargs
)
|