Skip to content

Instantly share code, notes, and snippets.

@padeoe
Last active February 27, 2025 16:25
Show Gist options
  • Save padeoe/697678ab8e528b85a2a7bddafea1fa4f to your computer and use it in GitHub Desktop.
Save padeoe/697678ab8e528b85a2a7bddafea1fa4f to your computer and use it in GitHub Desktop.
CLI-Tool for download Huggingface models and datasets with aria2/wget: hfd

🤗Huggingface Model Downloader

Note

(2025-01-08) Add feature for 🏷️Tag(Revision) Selection, contributed by @Bamboo-D.
(2024-12-17) Add feature for ⚡Quick Startup and ⏭️Fast Resume, enabling skipping of downloaded files, while removing the git clone dependency to accelerate file list retrieval.

Considering the lack of multi-threaded download support in the official huggingface-cli, and the inadequate error handling in hf_transfer, This command-line tool leverages curl and aria2c for fast and robust downloading of models and datasets.

Features

  • ⏯️ Resume from breakpoint: You can re-run it or Ctrl+C anytime.
  • 🚀 Multi-threaded Download: Utilize multiple threads to speed up the download process.
  • 🚫 File Exclusion: Use --exclude or --include to skip or specify files, save time for models with duplicate formats (e.g., *.bin or *.safetensors).
  • 🔐 Auth Support: For gated models that require Huggingface login, use --hf_username and --hf_token to authenticate.
  • 🪞 Mirror Site Support: Set up with HF_ENDPOINT environment variable.
  • 🌍 Proxy Support: Set up with https_proxy environment variable.
  • 📦 Simple: Minimal dependencies, requires only curl and wget, while aria2 and jq are optional for better performance.
  • 🏷️ Tag Selection: Support downloading specific model/dataset revision using --revision.

Usage

First, Download hfd.sh or clone this repo, and then grant execution permission to the script.

chmod a+x hfd.sh

you can create an alias for convenience

alias hfd="$PWD/hfd.sh"

Usage Instructions

$ ./hfd.sh --help
Usage:
  hfd <REPO_ID> [--include include_pattern1 include_pattern2 ...] [--exclude exclude_pattern1 exclude_pattern2 ...] [--hf_username username] [--hf_token token] [--tool aria2c|wget] [-x threads] [-j jobs] [--dataset] [--local-dir path] [--revision rev]

Description:
  Downloads a model or dataset from Hugging Face using the provided repo ID.

Arguments:
  REPO_ID         The Hugging Face repo ID (Required)
                  Format: 'org_name/repo_name' or legacy format (e.g., gpt2)
Options:
  include/exclude_pattern The patterns to match against file path, supports wildcard characters.
                  e.g., '--exclude *.safetensor *.md', '--include vae/*'.
  --include       (Optional) Patterns to include files for downloading (supports multiple patterns).
  --exclude       (Optional) Patterns to exclude files from downloading (supports multiple patterns).
  --hf_username   (Optional) Hugging Face username for authentication (not email).
  --hf_token      (Optional) Hugging Face token for authentication.
  --tool          (Optional) Download tool to use: aria2c (default) or wget.
  -x              (Optional) Number of download threads for aria2c (default: 4).
  -j              (Optional) Number of concurrent downloads for aria2c (default: 5).
  --dataset       (Optional) Flag to indicate downloading a dataset.
  --local-dir     (Optional) Directory path to store the downloaded data.
                             Defaults to the current directory with a subdirectory named 'repo_name'
                             if REPO_ID is is composed of 'org_name/repo_name'.
  --revision      (Optional) Model/Dataset revision to download (default: main).

Example:
  hfd gpt2
  hfd bigscience/bloom-560m --exclude *.bin *.msgpack onnx/*
  hfd meta-llama/Llama-2-7b --hf_username myuser --hf_token mytoken -x 4
  hfd lavita/medical-qa-shared-task-v1-toy --dataset
  hfd bartowski/Phi-3.5-mini-instruct-exl2 --revision 5_0

Download a model

hfd bigscience/bloom-560m

Download a model need login

Get huggingface token from https://huggingface.co/settings/tokens, then

hfd meta-llama/Llama-2-7b --hf_username YOUR_HF_USERNAME_NOT_EMAIL --hf_token YOUR_HF_TOKEN

Download a model and exclude certain files (e.g., .safetensors)

hfd bigscience/bloom-560m --exclude *.bin *.msgpack onnx/*

You can also exclude multiple pattern like that

hfd bigscience/bloom-560m --exclude *.bin --exclude *.msgpack --exclude onnx/*

Download specific files using include patterns

hfd Qwen/Qwen2.5-Coder-32B-Instruct-GGUF --include *q2_k*.gguf

Download a dataset

hfd lavita/medical-qa-shared-task-v1-toy --dataset

Download a specific revision of a model

hfd bartowski/Phi-3.5-mini-instruct-exl2 --revision 5_0

Multi-threading and Parallel Downloads

The script supports two types of parallelism when using aria2c:

  • Threads per File (-x): Controls connections per file, usage: hfd gpt2 -x 8, recommended: 4-8, default: 4 threads.

  • Concurrent Files (-j): Controls simultaneous file downloads, usage: hfd gpt2 -j 3, recommended: 3-8, default: 5 files.

Combined usage:

hfd gpt2 -x 8 -j 3  # 8 threads per file, 3 files at once
#!/usr/bin/env bash
# Color definitions
RED='\033[0;31m'; GREEN='\033[0;32m'; YELLOW='\033[1;33m'; NC='\033[0m' # No Color
trap 'printf "${YELLOW}\nDownload interrupted. You can resume by re-running the command.\n${NC}"; exit 1' INT
display_help() {
cat << EOF
Usage:
hfd <REPO_ID> [--include include_pattern1 include_pattern2 ...] [--exclude exclude_pattern1 exclude_pattern2 ...] [--hf_username username] [--hf_token token] [--tool aria2c|wget] [-x threads] [-j jobs] [--dataset] [--local-dir path] [--revision rev]
Description:
Downloads a model or dataset from Hugging Face using the provided repo ID.
Arguments:
REPO_ID The Hugging Face repo ID (Required)
Format: 'org_name/repo_name' or legacy format (e.g., gpt2)
Options:
include/exclude_pattern The patterns to match against file path, supports wildcard characters.
e.g., '--exclude *.safetensor *.md', '--include vae/*'.
--include (Optional) Patterns to include files for downloading (supports multiple patterns).
--exclude (Optional) Patterns to exclude files from downloading (supports multiple patterns).
--hf_username (Optional) Hugging Face username for authentication (not email).
--hf_token (Optional) Hugging Face token for authentication.
--tool (Optional) Download tool to use: aria2c (default) or wget.
-x (Optional) Number of download threads for aria2c (default: 4).
-j (Optional) Number of concurrent downloads for aria2c (default: 5).
--dataset (Optional) Flag to indicate downloading a dataset.
--local-dir (Optional) Directory path to store the downloaded data.
Defaults to the current directory with a subdirectory named 'repo_name'
if REPO_ID is is composed of 'org_name/repo_name'.
--revision (Optional) Model/Dataset revision to download (default: main).
Example:
hfd gpt2
hfd bigscience/bloom-560m --exclude *.safetensors
hfd meta-llama/Llama-2-7b --hf_username myuser --hf_token mytoken -x 4
hfd lavita/medical-qa-shared-task-v1-toy --dataset
hfd bartowski/Phi-3.5-mini-instruct-exl2 --revision 5_0
EOF
exit 1
}
[[ -z "$1" || "$1" =~ ^-h || "$1" =~ ^--help ]] && display_help
REPO_ID=$1
shift
# Default values
TOOL="aria2c"
THREADS=4
CONCURRENT=5
HF_ENDPOINT=${HF_ENDPOINT:-"https://huggingface.co"}
INCLUDE_PATTERNS=()
EXCLUDE_PATTERNS=()
REVISION="main"
validate_number() {
[[ "$2" =~ ^[1-9][0-9]*$ && "$2" -le "$3" ]] || { printf "${RED}[Error] $1 must be 1-$3${NC}\n"; exit 1; }
}
# Argument parsing
while [[ $# -gt 0 ]]; do
case $1 in
--include) shift; while [[ $# -gt 0 && ! ($1 =~ ^--) && ! ($1 =~ ^-[^-]) ]]; do INCLUDE_PATTERNS+=("$1"); shift; done ;;
--exclude) shift; while [[ $# -gt 0 && ! ($1 =~ ^--) && ! ($1 =~ ^-[^-]) ]]; do EXCLUDE_PATTERNS+=("$1"); shift; done ;;
--hf_username) HF_USERNAME="$2"; shift 2 ;;
--hf_token) HF_TOKEN="$2"; shift 2 ;;
--tool)
case $2 in
aria2c|wget)
TOOL="$2"
;;
*)
printf "%b[Error] Invalid tool. Use 'aria2c' or 'wget'.%b\n" "$RED" "$NC"
exit 1
;;
esac
shift 2
;;
-x) validate_number "threads (-x)" "$2" 10; THREADS="$2"; shift 2 ;;
-j) validate_number "concurrent downloads (-j)" "$2" 10; CONCURRENT="$2"; shift 2 ;;
--dataset) DATASET=1; shift ;;
--local-dir) LOCAL_DIR="$2"; shift 2 ;;
--revision) REVISION="$2"; shift 2 ;;
*) display_help ;;
esac
done
# Generate current command string
generate_command_string() {
local cmd_string="REPO_ID=$REPO_ID"
cmd_string+=" TOOL=$TOOL"
cmd_string+=" INCLUDE_PATTERNS=${INCLUDE_PATTERNS[*]}"
cmd_string+=" EXCLUDE_PATTERNS=${EXCLUDE_PATTERNS[*]}"
cmd_string+=" DATASET=${DATASET:-0}"
cmd_string+=" HF_USERNAME=${HF_USERNAME:-}"
cmd_string+=" HF_TOKEN=${HF_TOKEN:-}"
cmd_string+=" HF_TOKEN=${HF_ENDPOINT:-}"
cmd_string+=" REVISION=$REVISION"
echo "$cmd_string"
}
# Check if aria2, wget, curl are installed
check_command() {
if ! command -v $1 &>/dev/null; then
printf "%b%s is not installed. Please install it first.%b\n" "$RED" "$1" "$NC"
exit 1
fi
}
check_command curl; check_command "$TOOL"
LOCAL_DIR="${LOCAL_DIR:-${REPO_ID#*/}}"
mkdir -p "$LOCAL_DIR/.hfd"
if [[ "$DATASET" == 1 ]]; then
METADATA_API_PATH="datasets/$REPO_ID"
DOWNLOAD_API_PATH="datasets/$REPO_ID"
CUT_DIRS=5
else
METADATA_API_PATH="models/$REPO_ID"
DOWNLOAD_API_PATH="$REPO_ID"
CUT_DIRS=4
fi
# Modify API URL, construct based on revision
if [[ "$REVISION" != "main" ]]; then
METADATA_API_PATH="$METADATA_API_PATH/revision/$REVISION"
fi
API_URL="$HF_ENDPOINT/api/$METADATA_API_PATH"
METADATA_FILE="$LOCAL_DIR/.hfd/repo_metadata.json"
# Fetch and save metadata
fetch_and_save_metadata() {
status_code=$(curl -L -s -w "%{http_code}" -o "$METADATA_FILE" ${HF_TOKEN:+-H "Authorization: Bearer $HF_TOKEN"} "$API_URL")
RESPONSE=$(cat "$METADATA_FILE")
if [ "$status_code" -eq 200 ]; then
printf "%s\n" "$RESPONSE"
else
printf "%b[Error] Failed to fetch metadata from $API_URL. HTTP status code: $status_code.%b\n$RESPONSE\n" "${RED}" "${NC}" >&2
rm $METADATA_FILE
exit 1
fi
}
check_authentication() {
local response="$1"
if command -v jq &>/dev/null; then
local gated
gated=$(echo "$response" | jq -r '.gated // false')
if [[ "$gated" != "false" && ( -z "$HF_TOKEN" || -z "$HF_USERNAME" ) ]]; then
printf "${RED}The repository requires authentication, but --hf_username and --hf_token is not passed. Please get token from https://huggingface.co/settings/tokens.\nExiting.\n${NC}"
exit 1
fi
else
if echo "$response" | grep -q '"gated":[^f]' && [[ -z "$HF_TOKEN" || -z "$HF_USERNAME" ]]; then
printf "${RED}The repository requires authentication, but --hf_username and --hf_token is not passed. Please get token from https://huggingface.co/settings/tokens.\nExiting.\n${NC}"
exit 1
fi
fi
}
if [[ ! -f "$METADATA_FILE" ]]; then
printf "%bFetching repo metadata...%b\n" "$YELLOW" "$NC"
RESPONSE=$(fetch_and_save_metadata) || exit 1
check_authentication "$RESPONSE"
else
printf "%bUsing cached metadata: $METADATA_FILE%b\n" "$GREEN" "$NC"
RESPONSE=$(cat "$METADATA_FILE")
check_authentication "$RESPONSE"
fi
should_regenerate_filelist() {
local command_file="$LOCAL_DIR/.hfd/last_download_command"
local current_command=$(generate_command_string)
# If file list doesn't exist, regenerate
if [[ ! -f "$LOCAL_DIR/$fileslist_file" ]]; then
echo "$current_command" > "$command_file"
return 0
fi
# If command file doesn't exist, regenerate
if [[ ! -f "$command_file" ]]; then
echo "$current_command" > "$command_file"
return 0
fi
# Compare current command with saved command
local saved_command=$(cat "$command_file")
if [[ "$current_command" != "$saved_command" ]]; then
echo "$current_command" > "$command_file"
return 0
fi
return 1
}
fileslist_file=".hfd/${TOOL}_urls.txt"
if should_regenerate_filelist; then
# Remove existing file list if it exists
[[ -f "$LOCAL_DIR/$fileslist_file" ]] && rm "$LOCAL_DIR/$fileslist_file"
printf "%bGenerating file list...%b\n" "$YELLOW" "$NC"
# Convert include and exclude patterns to regex
INCLUDE_REGEX=""
EXCLUDE_REGEX=""
if ((${#INCLUDE_PATTERNS[@]})); then
INCLUDE_REGEX=$(printf '%s\n' "${INCLUDE_PATTERNS[@]}" | sed 's/\./\\./g; s/\*/.*/g' | paste -sd '|' -)
fi
if ((${#EXCLUDE_PATTERNS[@]})); then
EXCLUDE_REGEX=$(printf '%s\n' "${EXCLUDE_PATTERNS[@]}" | sed 's/\./\\./g; s/\*/.*/g' | paste -sd '|' -)
fi
# Check if jq is available
if command -v jq &>/dev/null; then
process_with_jq() {
if [[ "$TOOL" == "aria2c" ]]; then
printf "%s" "$RESPONSE" | jq -r \
--arg endpoint "$HF_ENDPOINT" \
--arg repo_id "$DOWNLOAD_API_PATH" \
--arg token "$HF_TOKEN" \
--arg include_regex "$INCLUDE_REGEX" \
--arg exclude_regex "$EXCLUDE_REGEX" \
--arg revision "$REVISION" \
'
.siblings[]
| select(
.rfilename != null
and ($include_regex == "" or (.rfilename | test($include_regex)))
and ($exclude_regex == "" or (.rfilename | test($exclude_regex) | not))
)
| [
($endpoint + "/" + $repo_id + "/resolve/" + $revision + "/" + .rfilename),
" dir=" + (.rfilename | split("/")[:-1] | join("/")),
" out=" + (.rfilename | split("/")[-1]),
if $token != "" then " header=Authorization: Bearer " + $token else empty end,
""
]
| join("\n")
'
else
printf "%s" "$RESPONSE" | jq -r \
--arg endpoint "$HF_ENDPOINT" \
--arg repo_id "$DOWNLOAD_API_PATH" \
--arg include_regex "$INCLUDE_REGEX" \
--arg exclude_regex "$EXCLUDE_REGEX" \
--arg revision "$REVISION" \
'
.siblings[]
| select(
.rfilename != null
and ($include_regex == "" or (.rfilename | test($include_regex)))
and ($exclude_regex == "" or (.rfilename | test($exclude_regex) | not))
)
| ($endpoint + "/" + $repo_id + "/resolve/" + $revision + "/" + .rfilename)
'
fi
}
result=$(process_with_jq)
printf "%s\n" "$result" > "$LOCAL_DIR/$fileslist_file"
else
printf "%b[Warning] jq not installed, using grep/awk for metadata json parsing (slower). Consider installing jq for better parsing performance.%b\n" "$YELLOW" "$NC"
process_with_grep_awk() {
local include_pattern=""
local exclude_pattern=""
local output=""
if ((${#INCLUDE_PATTERNS[@]})); then
include_pattern=$(printf '%s\n' "${INCLUDE_PATTERNS[@]}" | sed 's/\./\\./g; s/\*/.*/g' | paste -sd '|' -)
fi
if ((${#EXCLUDE_PATTERNS[@]})); then
exclude_pattern=$(printf '%s\n' "${EXCLUDE_PATTERNS[@]}" | sed 's/\./\\./g; s/\*/.*/g' | paste -sd '|' -)
fi
local files=$(printf '%s' "$RESPONSE" | grep -o '"rfilename":"[^"]*"' | awk -F'"' '{print $4}')
if [[ -n "$include_pattern" ]]; then
files=$(printf '%s\n' "$files" | grep -E "$include_pattern")
fi
if [[ -n "$exclude_pattern" ]]; then
files=$(printf '%s\n' "$files" | grep -vE "$exclude_pattern")
fi
while IFS= read -r file; do
if [[ -n "$file" ]]; then
if [[ "$TOOL" == "aria2c" ]]; then
output+="$HF_ENDPOINT/$DOWNLOAD_API_PATH/resolve/$REVISION/$file"$'\n'
output+=" dir=$(dirname "$file")"$'\n'
output+=" out=$(basename "$file")"$'\n'
[[ -n "$HF_TOKEN" ]] && output+=" header=Authorization: Bearer $HF_TOKEN"$'\n'
output+=$'\n'
else
output+="$HF_ENDPOINT/$DOWNLOAD_API_PATH/resolve/$REVISION/$file"$'\n'
fi
fi
done <<< "$files"
printf '%s' "$output"
}
result=$(process_with_grep_awk)
printf "%s\n" "$result" > "$LOCAL_DIR/$fileslist_file"
fi
else
printf "%bResume from file list: $LOCAL_DIR/$fileslist_file%b\n" "$GREEN" "$NC"
fi
# Perform download
printf "${YELLOW}Starting download with $TOOL to $LOCAL_DIR...\n${NC}"
cd "$LOCAL_DIR"
if [[ "$TOOL" == "aria2c" ]]; then
aria2c --console-log-level=error --file-allocation=none -x "$THREADS" -j "$CONCURRENT" -s "$THREADS" -k 1M -c -i "$fileslist_file" --save-session="$fileslist_file"
elif [[ "$TOOL" == "wget" ]]; then
wget -x -nH --cut-dirs="$CUT_DIRS" ${HF_TOKEN:+--header="Authorization: Bearer $HF_TOKEN"} --input-file="$fileslist_file" --continue
fi
if [[ $? -eq 0 ]]; then
printf "${GREEN}Download completed successfully. Repo directory: $PWD\n${NC}"
else
printf "${RED}Download encountered errors.\n${NC}"
exit 1
fi
@groklab
Copy link

groklab commented Feb 8, 2025

各位大佬,请教个初级问题。为啥aria2c几乎下载不动,wget还能稳定下载?
我运行./hfd.sh bigscience/bloom-560m --local-dir /models --tool wget 差不多7MB/s速度。这个wget是不是不能并行?
我运行./hfd.sh bigscience/bloom-560m --local-dir /models -x 1 -j 1 就下载几个MB之后就不动了。改变并行数量也是一样的行为。
环境是公司服务器。
请问咋回事?谢谢。

不太合理,用你的命令测试表现正常,没能复现

谢谢大佬。没有sudo, 我是从conda-forge里面下载的aria2cjq。不知道是不是这个导致的。再就是可能公司IT做了啥限制。。

更新:今天一早又试了下,aria2c模式又可以了!谢谢🙏

@CHN-STUDENT
Copy link

CHN-STUDENT commented Feb 9, 2025

@padeoe 话说佬,是不是最近负载过大,我遇到了403错误

[#7ce225 16GiB/46GiB(35%) CN:1 DL:14MiB ETA:35m34s]                                                                                                                                                                                                                       
02/09 03:20:14 [ERROR] CUID#12 - Download aborted. URI=https://hf-mirror.com/unsloth/DeepSeek-R1-GGUF/resolve/main/DeepSeek-R1-UD-IQ2_XXS/DeepSeek-R1-UD-IQ2_XXS-00002-of-00004.gguf
Exception: [AbstractCommand.cc:351] errorCode=22 URI=https://cdn-lfs-us-1.hf-mirror.com/repos/f3/c8/f3c8ca66be2bd64c06f26b308de2e2f3bfff2bf40b0c143e9f25f4c929ecfe57/ea1f482a3bde4e8303b52a689b2f7fa0bcfd6bdf311be7704dff6b824b03a5ed?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27DeepSeek-R1-UD-IQ2_XXS-00002-of-00004.gguf%3B+filename%3D%22DeepSeek-R1-UD-IQ2_XXS-00002-of-00004.gguf%22%3B&Expires=1739038493&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTczOTAzODQ5M319LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy11cy0xLmhmLmNvL3JlcG9zL2YzL2M4L2YzYzhjYTY2YmUyYmQ2NGMwNmYyNmIzMDhkZTJlMmYzYmZmZjJiZjQwYjBjMTQzZTlmMjVmNGM5MjllY2ZlNTcvZWExZjQ4MmEzYmRlNGU4MzAzYjUyYTY4OWIyZjdmYTBiY2ZkNmJkZjMxMWJlNzcwNGRmZjZiODI0YjAzYTVlZD9yZXNwb25zZS1jb250ZW50LWRpc3Bvc2l0aW9uPSoifV19&Signature=b-zZpWOIufCWgh4-RoHbVqiUj8KUIOTbO%7EY6Qpnl7%7E6NS6S8IzmhUbZ4nKITbBhI-W5sIUYcbVOhyjDKhtp2R6O2PLXkvl12rvobvG3ynibB7yI%7EKoP5CFJ7pEkLAG4XYpo4Gvhe7TI63MQcij%7EtnOkoeI6dSm7%7EUqlY9vrCxQ%7EiVtUuqu%7E1ZVq794HO8n4zF8LbGAwII27gMB4WhQpUzRj-j2AZ54BSG9ftpndsbsTcqtIz6kKXdzowv6O13pMbVEIgFiHActdLU-qwCkHTCzCujoG59x0EaS--ZQ6QIoe1dR0gTCRaSE2wlWFI9Q2kcK%7Eddf3WKWe50Cou86-H1g__&Key-Pair-Id=K24J24Z295AEI9
  -> [HttpSkipResponseCommand.cc:239] errorCode=22 响应状态不成功。状态=403

下载结果:
gid   |stat|avg speed  |path/URI
======+====+===========+=======================================================
b790d9|ERR |   653KiB/s|DeepSeek-R1-UD-IQ2_XXS/DeepSeek-R1-UD-IQ2_XXS-00004-of-00004.gguf
d6b953|ERR |   764KiB/s|DeepSeek-R1-UD-IQ2_XXS/DeepSeek-R1-UD-IQ2_XXS-00003-of-00004.gguf
e01a1c|ERR |   0.9MiB/s|DeepSeek-R1-UD-IQ2_XXS/DeepSeek-R1-UD-IQ2_XXS-00001-of-00004.gguf
7ce225|ERR |   679KiB/s|DeepSeek-R1-UD-IQ2_XXS/DeepSeek-R1-UD-IQ2_XXS-00002-of-00004.gguf

状态标识:
(ERR):发生错误。

传输重启后 aria2 可继续该下载。
如果发生任何错误,请参阅日志文件。详细信息见帮助/手册页面中的“-l”选项。
Download encountered errors.

更新:直接通过原命令继续下载,linux还是用的太少了,不太熟悉了。

话说能不能加个功能,下载失败继续重试下载,方便挂后台下载。

@marvelcell
Copy link

经排查是hf-mirror压力过大,导致的服务端问题,正在已经解决,(但近期由于DeepSeek相关模型下载量激增导致hf-mirror服务器较卡顿,可尝试备用端点)

请问大佬,备用端点的地址是什么呢?

@padeoe
Copy link
Author

padeoe commented Feb 10, 2025

经排查是hf-mirror压力过大,导致的服务端问题,正在已经解决,(但近期由于DeepSeek相关模型下载量激增导致hf-mirror服务器较卡顿,可尝试备用端点)

请问大佬,备用端点的地址是什么呢?

https://alpha.hf-mirror.com

可加网站首页微信群跟进备用地址

@Oklahomawhore
Copy link

换备用端点可以了,赞👍

经排查是hf-mirror压力过大,导致的服务端问题,正在已经解决,(但近期由于DeepSeek相关模型下载量激增导致hf-mirror服务器较卡顿,可尝试备用端点)

请问大佬,备用端点的地址是什么呢?

https://alpha.hf-mirror.com

可加网站首页微信群跟进备用地址

@xiuyanDL
Copy link

经排查是hf-mirror压力过大,导致的服务端问题,正在已经解决,(但近期由于DeepSeek相关模型下载量激增导致hf-mirror服务器较卡顿,可尝试备用端点)

请问大佬,备用端点的地址是什么呢?

https://alpha.hf-mirror.com

可加网站首页微信群跟进备用地址

这个alpha备用端点有个bug :
如果服务器本身开着代理的话, 端点配置为https://alpha.hf-mirror.com/的时候, 会走默认的cdn-lfs-us-1.hf.co,也就是会消耗代理的流量;
如果服务器本身没有代理,那么会走cdn-lfs-us-1.alpha.hf-mirror.com 。

但是老端点https://hf-mirror.com/没有上面所说的这个问题,一直走的是cdn-lfs-us-1.hf-mirror.com。

这个问题请麻烦看一下

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment