vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

patch: Apply fix for downloading related model files

Browse source
This commit is contained in:
teleprint-me 2024-05-08 18:50:28 -04:00
parent 17f2243d6c
commit de3d9e3b33
No known key found for this signature in database
GPG key ID: B0D11345E65C4D48
1 changed files with 4 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

19
convert-hf-to-gguf-update.py
View file

@ -102,8 +102,7 @@ for model in models:
repo = model["repo"]
tokt = model["tokt"]
# set url paths
url_main = f"{repo}/raw/main"
# NOTE: We should always be using resolve to download files
url_resolve = f"{repo}/resolve/main"
# set dir paths
@ -138,28 +137,18 @@ for model in models:
)
else: # Get the models tokenizer
download_file_with_auth(
url=f"{url_main}/tokenizer.json",
url=f"{url_resolve}/tokenizer.json",
token=token,
save_path=model_tokenizer_path
)
# Get the models hyper params
download_file_with_auth(
url=f"{url_main}/config.json",
url=f"{url_resolve}/config.json",
token=token,
save_path=f"{model_name_or_path}/config.json"
)
# if downloaded file is less than 1KB, we likely need to download an LFS instead
if os.path.getsize(model_tokenizer_path) < 1024:
# remove the file
os.remove(model_tokenizer_path)
download_file_with_auth(
url=f"{url_resolve}/tokenizer.json",
token=token,
save_path=model_tokenizer_path
)
# Handle sentencepiece tokenizer
if tokt == TOKENIZER_TYPE.SPM:
download_file_with_auth(
@ -170,7 +159,7 @@ for model in models:
# Get the tokenizer config
download_file_with_auth(
url=f"{url_main}/tokenizer_config.json",
url=f"{url_resolve}/tokenizer_config.json",
token=token,
save_path=f"{model_name_or_path}/tokenizer_config.json"
)