2023-01-06 23:57:31 +01:00
|
|
|
'''
|
|
|
|
Downloads models from Hugging Face to models/model-name.
|
|
|
|
|
|
|
|
Example:
|
2023-04-09 22:00:59 +02:00
|
|
|
python download-model.py facebook/opt-1.3b
|
2023-01-06 23:57:31 +01:00
|
|
|
|
|
|
|
'''
|
2023-03-10 04:41:10 +01:00
|
|
|
|
2023-02-10 19:40:03 +01:00
|
|
|
import argparse
|
2023-03-10 04:41:10 +01:00
|
|
|
import base64
|
2023-03-30 01:26:44 +02:00
|
|
|
import datetime
|
2023-03-31 06:31:47 +02:00
|
|
|
import hashlib
|
2023-02-24 18:06:42 +01:00
|
|
|
import json
|
2023-02-10 19:40:03 +01:00
|
|
|
import re
|
2023-06-01 05:11:21 +02:00
|
|
|
import os
|
2023-01-20 21:51:56 +01:00
|
|
|
import sys
|
2023-01-07 20:33:43 +01:00
|
|
|
from pathlib import Path
|
2023-02-10 19:40:03 +01:00
|
|
|
|
|
|
|
import requests
|
|
|
|
import tqdm
|
2023-03-29 03:29:20 +02:00
|
|
|
from tqdm.contrib.concurrent import thread_map
|
2023-01-20 21:51:56 +01:00
|
|
|
|
2023-04-09 21:59:59 +02:00
|
|
|
|
2023-02-17 03:04:13 +01:00
|
|
|
def select_model_from_default_options():
|
|
|
|
models = {
|
2023-04-01 19:47:47 +02:00
|
|
|
"OPT 6.7B": ("facebook", "opt-6.7b", "main"),
|
|
|
|
"OPT 2.7B": ("facebook", "opt-2.7b", "main"),
|
|
|
|
"OPT 1.3B": ("facebook", "opt-1.3b", "main"),
|
|
|
|
"OPT 350M": ("facebook", "opt-350m", "main"),
|
|
|
|
"GALACTICA 6.7B": ("facebook", "galactica-6.7b", "main"),
|
|
|
|
"GALACTICA 1.3B": ("facebook", "galactica-1.3b", "main"),
|
|
|
|
"GALACTICA 125M": ("facebook", "galactica-125m", "main"),
|
|
|
|
"Pythia-6.9B-deduped": ("EleutherAI", "pythia-6.9b-deduped", "main"),
|
|
|
|
"Pythia-2.8B-deduped": ("EleutherAI", "pythia-2.8b-deduped", "main"),
|
|
|
|
"Pythia-1.4B-deduped": ("EleutherAI", "pythia-1.4b-deduped", "main"),
|
|
|
|
"Pythia-410M-deduped": ("EleutherAI", "pythia-410m-deduped", "main"),
|
2023-02-17 03:04:13 +01:00
|
|
|
}
|
|
|
|
|
2023-05-11 00:00:14 +02:00
|
|
|
choices = {}
|
2023-02-17 03:04:13 +01:00
|
|
|
print("Select the model that you want to download:\n")
|
2023-04-07 05:15:45 +02:00
|
|
|
for i, name in enumerate(models):
|
|
|
|
char = chr(ord('A') + i)
|
2023-02-17 03:04:13 +01:00
|
|
|
choices[char] = name
|
|
|
|
print(f"{char}) {name}")
|
2023-05-11 00:00:14 +02:00
|
|
|
|
2023-04-21 17:54:50 +02:00
|
|
|
char_hugging = chr(ord('A') + len(models))
|
|
|
|
print(f"{char_hugging}) Manually specify a Hugging Face model")
|
|
|
|
char_exit = chr(ord('A') + len(models) + 1)
|
|
|
|
print(f"{char_exit}) Do not download a model")
|
2023-02-17 03:04:13 +01:00
|
|
|
print()
|
|
|
|
print("Input> ", end='')
|
2023-02-20 19:50:48 +01:00
|
|
|
choice = input()[0].strip().upper()
|
2023-04-21 17:54:50 +02:00
|
|
|
if choice == char_exit:
|
|
|
|
exit()
|
|
|
|
elif choice == char_hugging:
|
2023-05-11 00:00:14 +02:00
|
|
|
print("""\nType the name of your desired Hugging Face model in the format organization/name.
|
2023-02-17 03:04:13 +01:00
|
|
|
|
|
|
|
Examples:
|
|
|
|
facebook/opt-1.3b
|
2023-04-01 20:03:24 +02:00
|
|
|
EleutherAI/pythia-1.4b-deduped
|
2023-02-17 03:04:13 +01:00
|
|
|
""")
|
|
|
|
|
|
|
|
print("Input> ", end='')
|
|
|
|
model = input()
|
|
|
|
branch = "main"
|
|
|
|
else:
|
|
|
|
arr = models[choices[choice]]
|
|
|
|
model = f"{arr[0]}/{arr[1]}"
|
|
|
|
branch = arr[2]
|
|
|
|
|
|
|
|
return model, branch
|
|
|
|
|
2023-04-07 05:15:45 +02:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
class ModelDownloader:
|
|
|
|
def __init__(self):
|
|
|
|
self.s = requests.Session()
|
|
|
|
if os.getenv('HF_USER') is not None and os.getenv('HF_PASS') is not None:
|
|
|
|
self.s.auth = (os.getenv('HF_USER'), os.getenv('HF_PASS'))
|
|
|
|
|
|
|
|
def sanitize_model_and_branch_names(self, model, branch):
|
|
|
|
if model[-1] == '/':
|
|
|
|
model = model[:-1]
|
|
|
|
|
|
|
|
if branch is None:
|
|
|
|
branch = "main"
|
|
|
|
else:
|
|
|
|
pattern = re.compile(r"^[a-zA-Z0-9._-]+$")
|
|
|
|
if not pattern.match(branch):
|
|
|
|
raise ValueError(
|
|
|
|
"Invalid branch name. Only alphanumeric characters, period, underscore and dash are allowed.")
|
|
|
|
|
|
|
|
return model, branch
|
|
|
|
|
|
|
|
def get_download_links_from_huggingface(self, model, branch, text_only=False):
|
|
|
|
base = "https://huggingface.co"
|
|
|
|
page = f"/api/models/{model}/tree/{branch}"
|
|
|
|
cursor = b""
|
|
|
|
|
|
|
|
links = []
|
|
|
|
sha256 = []
|
|
|
|
classifications = []
|
|
|
|
has_pytorch = False
|
|
|
|
has_pt = False
|
2023-06-06 12:05:32 +02:00
|
|
|
# has_ggml = False
|
2023-06-01 05:11:21 +02:00
|
|
|
has_safetensors = False
|
|
|
|
is_lora = False
|
|
|
|
while True:
|
|
|
|
url = f"{base}{page}" + (f"?cursor={cursor.decode()}" if cursor else "")
|
2023-06-08 16:16:38 +02:00
|
|
|
r = self.s.get(url, timeout=20)
|
2023-06-01 05:11:21 +02:00
|
|
|
r.raise_for_status()
|
|
|
|
content = r.content
|
|
|
|
|
|
|
|
dict = json.loads(content)
|
|
|
|
if len(dict) == 0:
|
|
|
|
break
|
|
|
|
|
|
|
|
for i in range(len(dict)):
|
|
|
|
fname = dict[i]['path']
|
|
|
|
if not is_lora and fname.endswith(('adapter_config.json', 'adapter_model.bin')):
|
|
|
|
is_lora = True
|
|
|
|
|
|
|
|
is_pytorch = re.match("(pytorch|adapter)_model.*\.bin", fname)
|
|
|
|
is_safetensors = re.match(".*\.safetensors", fname)
|
|
|
|
is_pt = re.match(".*\.pt", fname)
|
|
|
|
is_ggml = re.match(".*ggml.*\.bin", fname)
|
|
|
|
is_tokenizer = re.match("(tokenizer|ice).*\.model", fname)
|
|
|
|
is_text = re.match(".*\.(txt|json|py|md)", fname) or is_tokenizer
|
|
|
|
if any((is_pytorch, is_safetensors, is_pt, is_ggml, is_tokenizer, is_text)):
|
|
|
|
if 'lfs' in dict[i]:
|
|
|
|
sha256.append([fname, dict[i]['lfs']['oid']])
|
|
|
|
|
|
|
|
if is_text:
|
|
|
|
links.append(f"https://huggingface.co/{model}/resolve/{branch}/{fname}")
|
|
|
|
classifications.append('text')
|
|
|
|
continue
|
|
|
|
|
|
|
|
if not text_only:
|
|
|
|
links.append(f"https://huggingface.co/{model}/resolve/{branch}/{fname}")
|
|
|
|
if is_safetensors:
|
|
|
|
has_safetensors = True
|
|
|
|
classifications.append('safetensors')
|
|
|
|
elif is_pytorch:
|
|
|
|
has_pytorch = True
|
|
|
|
classifications.append('pytorch')
|
|
|
|
elif is_pt:
|
|
|
|
has_pt = True
|
|
|
|
classifications.append('pt')
|
|
|
|
elif is_ggml:
|
2023-06-06 12:05:32 +02:00
|
|
|
# has_ggml = True
|
2023-06-01 05:11:21 +02:00
|
|
|
classifications.append('ggml')
|
|
|
|
|
|
|
|
cursor = base64.b64encode(f'{{"file_name":"{dict[-1]["path"]}"}}'.encode()) + b':50'
|
|
|
|
cursor = base64.b64encode(cursor)
|
|
|
|
cursor = cursor.replace(b'=', b'%3D')
|
|
|
|
|
|
|
|
# If both pytorch and safetensors are available, download safetensors only
|
|
|
|
if (has_pytorch or has_pt) and has_safetensors:
|
|
|
|
for i in range(len(classifications) - 1, -1, -1):
|
|
|
|
if classifications[i] in ['pytorch', 'pt']:
|
|
|
|
links.pop(i)
|
|
|
|
|
|
|
|
return links, sha256, is_lora
|
|
|
|
|
|
|
|
def get_output_folder(self, model, branch, is_lora, base_folder=None):
|
|
|
|
if base_folder is None:
|
|
|
|
base_folder = 'models' if not is_lora else 'loras'
|
|
|
|
|
|
|
|
output_folder = f"{'_'.join(model.split('/')[-2:])}"
|
|
|
|
if branch != 'main':
|
|
|
|
output_folder += f'_{branch}'
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
output_folder = Path(base_folder) / output_folder
|
|
|
|
return output_folder
|
|
|
|
|
|
|
|
def get_single_file(self, url, output_folder, start_from_scratch=False):
|
|
|
|
filename = Path(url.rsplit('/', 1)[1])
|
|
|
|
output_path = output_folder / filename
|
2023-06-21 04:14:18 +02:00
|
|
|
headers = {}
|
|
|
|
mode = 'wb'
|
2023-06-01 05:11:21 +02:00
|
|
|
if output_path.exists() and not start_from_scratch:
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
# Check if the file has already been downloaded completely
|
2023-06-08 16:16:38 +02:00
|
|
|
r = self.s.get(url, stream=True, timeout=20)
|
2023-06-01 05:11:21 +02:00
|
|
|
total_size = int(r.headers.get('content-length', 0))
|
|
|
|
if output_path.stat().st_size >= total_size:
|
|
|
|
return
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
# Otherwise, resume the download from where it left off
|
|
|
|
headers = {'Range': f'bytes={output_path.stat().st_size}-'}
|
|
|
|
mode = 'ab'
|
|
|
|
|
2023-06-21 04:14:18 +02:00
|
|
|
with self.s.get(url, stream=True, headers=headers, timeout=20) as r:
|
|
|
|
r.raise_for_status() # Do not continue the download if the request was unsuccessful
|
2023-06-01 05:11:21 +02:00
|
|
|
total_size = int(r.headers.get('content-length', 0))
|
2023-06-21 04:14:18 +02:00
|
|
|
block_size = 1024 * 1024 # 1MB
|
|
|
|
with open(output_path, mode) as f:
|
|
|
|
with tqdm.tqdm(total=total_size,
|
|
|
|
unit='iB',
|
|
|
|
unit_scale=True,
|
|
|
|
bar_format='{l_bar}{bar}| {n_fmt:6}/{total_fmt:6} {rate_fmt:6}'
|
|
|
|
) as t:
|
|
|
|
count = 0
|
|
|
|
for data in r.iter_content(block_size):
|
|
|
|
t.update(len(data))
|
|
|
|
f.write(data)
|
|
|
|
if self.progress_bar is not None:
|
|
|
|
count += len(data)
|
|
|
|
self.progress_bar(float(count) / float(total_size), f"Downloading {filename}")
|
2023-06-01 05:11:21 +02:00
|
|
|
|
|
|
|
def start_download_threads(self, file_list, output_folder, start_from_scratch=False, threads=1):
|
|
|
|
thread_map(lambda url: self.get_single_file(url, output_folder, start_from_scratch=start_from_scratch), file_list, max_workers=threads, disable=True)
|
|
|
|
|
2023-06-21 04:14:18 +02:00
|
|
|
def download_model_files(self, model, branch, links, sha256, output_folder, progress_bar=None, start_from_scratch=False, threads=1):
|
2023-06-21 03:59:14 +02:00
|
|
|
self.progress_bar = progress_bar
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
# Creating the folder and writing the metadata
|
2023-06-21 04:14:18 +02:00
|
|
|
output_folder.mkdir(parents=True, exist_ok=True)
|
|
|
|
metadata = f'url: https://huggingface.co/{model}\n' \
|
|
|
|
f'branch: {branch}\n' \
|
|
|
|
f'download date: {datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")}\n'
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-21 04:14:18 +02:00
|
|
|
sha256_str = '\n'.join([f' {item[1]} {item[0]}' for item in sha256])
|
|
|
|
if sha256_str:
|
|
|
|
metadata += f'sha256sum:\n{sha256_str}'
|
2023-06-21 04:25:58 +02:00
|
|
|
|
2023-06-21 04:14:18 +02:00
|
|
|
metadata += '\n'
|
|
|
|
(output_folder / 'huggingface-metadata.txt').write_text(metadata)
|
2023-06-01 05:11:21 +02:00
|
|
|
|
|
|
|
# Downloading the files
|
|
|
|
print(f"Downloading the model to {output_folder}")
|
|
|
|
self.start_download_threads(links, output_folder, start_from_scratch=start_from_scratch, threads=threads)
|
|
|
|
|
|
|
|
def check_model_files(self, model, branch, links, sha256, output_folder):
|
|
|
|
# Validate the checksums
|
|
|
|
validated = True
|
2023-04-09 21:59:59 +02:00
|
|
|
for i in range(len(sha256)):
|
2023-06-01 05:11:21 +02:00
|
|
|
fpath = (output_folder / sha256[i][0])
|
|
|
|
|
|
|
|
if not fpath.exists():
|
|
|
|
print(f"The following file is missing: {fpath}")
|
2023-03-31 06:31:47 +02:00
|
|
|
validated = False
|
2023-06-01 05:11:21 +02:00
|
|
|
continue
|
2023-02-24 18:06:42 +01:00
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
with open(output_folder / sha256[i][0], "rb") as f:
|
|
|
|
bytes = f.read()
|
|
|
|
file_hash = hashlib.sha256(bytes).hexdigest()
|
|
|
|
if file_hash != sha256[i][1]:
|
|
|
|
print(f'Checksum failed: {sha256[i][0]} {sha256[i][1]}')
|
|
|
|
validated = False
|
|
|
|
else:
|
|
|
|
print(f'Checksum validated: {sha256[i][0]} {sha256[i][1]}')
|
|
|
|
|
|
|
|
if validated:
|
|
|
|
print('[+] Validated checksums of all model files!')
|
|
|
|
else:
|
|
|
|
print('[-] Invalid checksums. Rerun download-model.py with the --clean flag.')
|
2023-04-01 03:52:52 +02:00
|
|
|
|
2023-04-09 21:59:59 +02:00
|
|
|
|
|
|
|
if __name__ == '__main__':
|
2023-04-10 16:36:39 +02:00
|
|
|
|
|
|
|
parser = argparse.ArgumentParser()
|
|
|
|
parser.add_argument('MODEL', type=str, default=None, nargs='?')
|
|
|
|
parser.add_argument('--branch', type=str, default='main', help='Name of the Git branch to download from.')
|
|
|
|
parser.add_argument('--threads', type=int, default=1, help='Number of files to download simultaneously.')
|
|
|
|
parser.add_argument('--text-only', action='store_true', help='Only download text files (txt/json).')
|
|
|
|
parser.add_argument('--output', type=str, default=None, help='The folder where the model should be saved.')
|
|
|
|
parser.add_argument('--clean', action='store_true', help='Does not resume the previous download.')
|
|
|
|
parser.add_argument('--check', action='store_true', help='Validates the checksums of model files.')
|
|
|
|
args = parser.parse_args()
|
|
|
|
|
2023-04-09 21:59:59 +02:00
|
|
|
branch = args.branch
|
|
|
|
model = args.MODEL
|
|
|
|
if model is None:
|
|
|
|
model, branch = select_model_from_default_options()
|
|
|
|
|
2023-06-01 05:11:21 +02:00
|
|
|
downloader = ModelDownloader()
|
2023-04-09 21:59:59 +02:00
|
|
|
# Cleaning up the model/branch names
|
|
|
|
try:
|
2023-06-01 05:11:21 +02:00
|
|
|
model, branch = downloader.sanitize_model_and_branch_names(model, branch)
|
2023-04-09 21:59:59 +02:00
|
|
|
except ValueError as err_branch:
|
|
|
|
print(f"Error: {err_branch}")
|
|
|
|
sys.exit()
|
|
|
|
|
|
|
|
# Getting the download links from Hugging Face
|
2023-06-01 05:11:21 +02:00
|
|
|
links, sha256, is_lora = downloader.get_download_links_from_huggingface(model, branch, text_only=args.text_only)
|
2023-04-09 21:59:59 +02:00
|
|
|
|
|
|
|
# Getting the output folder
|
2023-06-01 05:11:21 +02:00
|
|
|
output_folder = downloader.get_output_folder(model, branch, is_lora, base_folder=args.output)
|
2023-04-09 21:59:59 +02:00
|
|
|
|
|
|
|
if args.check:
|
|
|
|
# Check previously downloaded files
|
2023-06-01 05:11:21 +02:00
|
|
|
downloader.check_model_files(model, branch, links, sha256, output_folder)
|
2023-04-09 21:59:59 +02:00
|
|
|
else:
|
|
|
|
# Download files
|
2023-06-01 05:11:21 +02:00
|
|
|
downloader.download_model_files(model, branch, links, sha256, output_folder, threads=args.threads)
|