|
#!/usr/bin/env bash |
|
RED='\033[0;31m'; GREEN='\033[0;32m'; YELLOW='\033[1;33m'; DIM='\033[2m'; BOLD='\033[1m'; NC='\033[0m' # No Color |
|
|
|
trap 'printf "\n%bInterrupted. Re-run to resume.%b\n" "$YELLOW" "$NC"; exit 130' INT |
|
|
|
# Format a byte count as a human-readable string (decimal units, matching the HF UI). |
|
human() { |
|
awk -v b="${1:-0}" 'BEGIN{u="B KB MB GB TB PB";n=split(u,a," ");i=1; |
|
while(b>=1000&&i<n){b/=1000;i++} printf (i==1?"%d%s":"%.2f%s"),b,a[i]}' |
|
} |
|
|
|
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 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 "$3" ]] || { printf "%b[Error] %s must be 1-%s%b\n" "$RED" "$1" "$3" "$NC"; 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) |
|
[[ "$2" == aria2c || "$2" == wget ]] || { printf "%b[Error] Invalid tool. Use 'aria2c' or 'wget'.%b\n" "$RED" "$NC"; exit 1; } |
|
TOOL="$2"; 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="$2"; shift 2 ;; |
|
*) display_help ;; |
|
esac |
|
done |
|
|
|
# A fingerprint of the options that affect the file list; a change forces regeneration. |
|
generate_command_string() { |
|
printf 'REPO_ID=%s TOOL=%s INCLUDE=%s EXCLUDE=%s DATASET=%s HF_USERNAME=%s HF_TOKEN=%s HF_ENDPOINT=%s REVISION=%s' \ |
|
"$REPO_ID" "$TOOL" "${INCLUDE_PATTERNS[*]}" "${EXCLUDE_PATTERNS[*]}" "${DATASET:-0}" \ |
|
"${HF_USERNAME:-}" "${HF_TOKEN:-}" "${HF_ENDPOINT:-}" "$REVISION" |
|
} |
|
|
|
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" |
|
|
|
REPO_API_PATH="models/$REPO_ID"; DOWNLOAD_API_PATH="$REPO_ID" |
|
[[ "$DATASET" == 1 ]] && { REPO_API_PATH="datasets/$REPO_ID"; DOWNLOAD_API_PATH=$REPO_API_PATH; } |
|
|
|
# wget --cut-dirs strips "<download_api_path>/resolve/<revision>/"; a fixed value wrongly |
|
# assumes the org/name form and eats a directory level for legacy (single-name) repo ids. |
|
CUT_DIRS=$(( $(printf '%s' "$DOWNLOAD_API_PATH" | tr -cd '/' | wc -c) + 3 )) |
|
|
|
# Metadata API URL (used for the gated/auth check); append revision when not main. |
|
METADATA_API_PATH="$REPO_API_PATH" |
|
[[ "$REVISION" != "main" ]] && METADATA_API_PATH="$METADATA_API_PATH/revision/$REVISION" |
|
API_URL="$HF_ENDPOINT/api/$METADATA_API_PATH?blobs=true" |
|
|
|
METADATA_FILE="$LOCAL_DIR/.hfd/repo_metadata.json" |
|
|
|
fetch_and_save_metadata() { |
|
status_code=$(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 %s. HTTP status code: %s.%b\n%s\n" "$RED" "$API_URL" "$status_code" "$NC" "$RESPONSE" >&2 |
|
rm -f "$METADATA_FILE" |
|
exit 1 |
|
fi |
|
} |
|
|
|
# Exit early if the repo is gated/private but no credentials were supplied. |
|
check_authentication() { |
|
local gated |
|
if command -v jq &>/dev/null; then |
|
gated=$(printf '%s' "$1" | jq -r '.gated // false') |
|
else |
|
printf '%s' "$1" | grep -q '"gated":[^f]' && gated=true || gated=false |
|
fi |
|
if [[ "$gated" != "false" && ( -z "$HF_TOKEN" || -z "$HF_USERNAME" ) ]]; then |
|
printf "%bThe repository requires authentication, but --hf_username and --hf_token is not passed. Please get token from https://huggingface.co/settings/tokens.\nExiting.\n%b" "$RED" "$NC" |
|
exit 1 |
|
fi |
|
} |
|
|
|
printf "%b%s%b (%s)\n" "$BOLD" "$REPO_ID" "$NC" "$REVISION" |
|
if [[ ! -f "$METADATA_FILE" ]]; then |
|
printf "%bFetching metadata...%b\n" "$DIM" "$NC" |
|
RESPONSE=$(fetch_and_save_metadata) || exit 1 |
|
else |
|
RESPONSE=$(cat "$METADATA_FILE") |
|
fi |
|
check_authentication "$RESPONSE" |
|
|
|
# Total bytes of the revision (cheap, branch-specific); empty if the endpoint lacks this API. |
|
fetch_treesize() { |
|
local json |
|
json=$(curl -sSL ${HF_TOKEN:+-H "Authorization: Bearer $HF_TOKEN"} "$HF_ENDPOINT/api/$REPO_API_PATH/treesize/$REVISION") || return |
|
if command -v jq &>/dev/null; then |
|
printf '%s' "$json" | jq -r '.size // empty' 2>/dev/null |
|
else |
|
printf '%s' "$json" | grep -o '"size":[0-9]*' | head -1 | grep -o '[0-9]*' |
|
fi |
|
} |
|
# Reuse the cached list only if the command is unchanged AND the manifest is intact (its line |
|
# count equals repo_info's total). An interrupted/failed walk leaves no repo_info, so it re-lists. |
|
should_regenerate_filelist() { |
|
local cmd="$LOCAL_DIR/.hfd/last_download_command" mf="$LOCAL_DIR/.hfd/manifest" info="$LOCAL_DIR/.hfd/repo_info" |
|
[[ -f "$mf" && -f "$info" && "$(generate_command_string)" == "$(cat "$cmd" 2>/dev/null)" \ |
|
&& "$(wc -l < "$mf")" == "$(cut -d' ' -f1 "$info" 2>/dev/null)" ]] && return 1 |
|
return 0 |
|
} |
|
|
|
fileslist_file=".hfd/${TOOL}_urls.txt" |
|
|
|
# Convert a list of wildcard patterns into a single alternation regex. |
|
patterns_to_regex() { |
|
(($#)) || return 0 |
|
printf '%s\n' "$@" | sed 's/\./\\./g; s/\*/.*/g' | paste -sd '|' - |
|
} |
|
|
|
# Emit "size<TAB>path" for every file in a tree page (jq, or a grep/awk fallback). |
|
emit_page_files() { |
|
if command -v jq &>/dev/null; then |
|
jq -r '.[] | select(.type=="file") | "\(.size)\t\(.path)"' "$1" |
|
else |
|
# Strip the nested lfs{} object (it has its own "size") so type/size/path align per entry. |
|
sed 's/"lfs":{[^}]*}//g' "$1" \ |
|
| grep -oE '"type":"[^"]*"|"size":[0-9]+|"path":"[^"]*"' \ |
|
| sed 's/"type":"//; s/"size"://; s/"path":"//; s/"$//' \ |
|
| awk 'NR%3==1{t=$0} NR%3==2{s=$0} NR%3==0{if(t=="file")print s"\t"$0}' |
|
fi |
|
} |
|
|
|
# Emit "size<TAB>path" for every file from the cached metadata siblings[] (needs jq). |
|
emit_siblings() { |
|
jq -r '(.siblings // [])[] | "\(.size // 0)\t\(.rfilename)"' "$METADATA_FILE" |
|
} |
|
|
|
# True when siblings[] is the complete list: its sizes sum to treesize (so it wasn't truncated, as |
|
# it is for very large repos). Lets us skip the slow tree walk for typical repos via one API call. |
|
siblings_complete() { |
|
command -v jq &>/dev/null && [[ "$REPO_SIZE" =~ ^[0-9]+$ ]] || return 1 |
|
local n s |
|
read -r n s < <(jq -r '(.siblings // []) as $s | "\($s|length) \([$s[].size // 0]|add // 0)"' "$METADATA_FILE" 2>/dev/null) |
|
(( ${n:-0} > 0 )) && [[ "${s:-0}" == "$REPO_SIZE" ]] |
|
} |
|
|
|
# Keep only "size<TAB>path" stdin lines matching include/exclude (the *_REGEX globals). |
|
filter_size_path() { |
|
local size path |
|
while IFS=$'\t' read -r size path; do |
|
[[ -z "$path" ]] && continue |
|
[[ -n "$INCLUDE_REGEX" && ! "$path" =~ $INCLUDE_REGEX ]] && continue |
|
[[ -n "$EXCLUDE_REGEX" && "$path" =~ $EXCLUDE_REGEX ]] && continue |
|
printf '%s\t%s\n' "${size:-0}" "$path" |
|
done |
|
} |
|
|
|
# Resumably walk the recursive tree into .hfd/manifest.partial (filtered size<TAB>path). After each |
|
# page, checkpoint .hfd/list_state (fingerprint / entries scanned / next cursor) so an interrupted or |
|
# failed walk continues from there instead of restarting. Progress (entries scanned) goes to stderr. |
|
walk_tree() { |
|
local state="$LOCAL_DIR/.hfd/list_state" partial="$LOCAL_DIR/.hfd/manifest.partial" |
|
local page="$LOCAL_DIR/.hfd/tree_page.json" fp url scanned=0 headers n saved_fp="" |
|
fp=$(generate_command_string) |
|
[[ -f "$state" && -f "$partial" ]] && { read -r saved_fp; read -r scanned; read -r url; } < "$state" |
|
if [[ "$saved_fp" != "$fp" ]]; then # no checkpoint, or it belongs to a different command |
|
url="$HF_ENDPOINT/api/$REPO_API_PATH/tree/$REVISION?recursive=true&expand=false" |
|
scanned=0; : > "$partial" |
|
elif [[ -s "$partial" && -n "$(tail -c1 "$partial")" ]]; then |
|
sed -i '$d' "$partial" # drop a torn final line from a kill mid-append (the page re-fetches) |
|
fi |
|
while [[ -n "$url" ]]; do |
|
headers=$(curl -sSL ${HF_TOKEN:+-H "Authorization: Bearer $HF_TOKEN"} -D - -o "$page" "$url") || return 1 |
|
if command -v jq &>/dev/null; then n=$(jq 'length' "$page" 2>/dev/null); else n=$(grep -o '"type":"' "$page" | wc -l); fi |
|
emit_page_files "$page" | filter_size_path >> "$partial" |
|
scanned=$(( scanned + ${n:-0} )) |
|
printf '\r\033[K%bListing files... %d scanned%b' "$DIM" "$scanned" "$NC" >&2 |
|
url=$(printf '%s' "$headers" | tr -d '\r' | grep -i '^link:' \ |
|
| grep -o '<[^>]*>;[[:space:]]*rel="next"' | sed -n 's/^<\([^>]*\)>.*/\1/p') |
|
printf '%s\n%s\n%s\n' "$fp" "$scanned" "$url" > "$state" # checkpoint after the page is in |
|
done |
|
return 0 |
|
} |
|
|
|
# Dedup .hfd/manifest.partial by path (a resumed walk may re-list its boundary page) into the final |
|
# manifest, and write "<count> <bytes>" totals to filelist_stats. |
|
finalize_filelist() { |
|
local mf="$LOCAL_DIR/.hfd/manifest" |
|
: > "$mf" |
|
awk -F'\t' -v mf="$mf" '!seen[$2]++ { print >> mf; n++; s+=$1 } END { print (n+0)" "(s+0) }' \ |
|
"$LOCAL_DIR/.hfd/manifest.partial" > "$LOCAL_DIR/.hfd/filelist_stats" |
|
} |
|
|
|
# Build tool input from the manifest; set NEED_COUNT (files still to fetch). One find pass diffs |
|
# the tree, keeping files missing or the wrong size (.aria2 sidecar = in-progress, kept): cheap |
|
# for huge repos, correct after a manual delete. Run from the local dir (manifest paths relative). |
|
build_download_list() { |
|
local needed=.hfd/needed size path dir cur |
|
if [[ ! -s .hfd/manifest ]]; then NEED_COUNT=0; : > "$fileslist_file"; return; fi |
|
awk -F'\t' ' |
|
FNR==NR { want[$2]=$1; ord[++n]=$2; next } |
|
{ p=$2; sub(/^\.\//,"",p); |
|
if (p ~ /\.aria2$/) { sub(/\.aria2$/,"",p); part[p]=1 } else have[p]=$1 } |
|
END { for (i=1;i<=n;i++) { q=ord[i]; |
|
if ((q in have) && have[q]==want[q] && !(q in part)) continue; |
|
print want[q] "\t" q } } |
|
' .hfd/manifest <(find . -type f ! -path './.hfd/*' -printf '%s\t%p\n' 2>/dev/null) > "$needed" |
|
NEED_COUNT=$(wc -l < "$needed") |
|
# Per needed file: drop a wrong-size copy -c/--continue can't fix in place (aria2c with no |
|
# .aria2 control file, or a wget file larger than expected) for a fresh refetch; emit the record. |
|
while IFS=$'\t' read -r size path; do |
|
[[ -z "$path" ]] && continue |
|
if [[ "$TOOL" == "aria2c" ]]; then |
|
[[ -e "$path" && ! -e "$path.aria2" ]] && rm -f "$path" |
|
dir="${path%/*}"; [[ "$dir" == "$path" ]] && dir="" |
|
printf '%s/%s/resolve/%s/%s\n dir=%s\n out=%s\n' "$HF_ENDPOINT" "$DOWNLOAD_API_PATH" "$REVISION" "$path" "$dir" "${path##*/}" |
|
[[ -n "$HF_TOKEN" ]] && printf ' header=Authorization: Bearer %s\n' "$HF_TOKEN" |
|
printf '\n' |
|
else |
|
cur=$(stat -c%s "$path" 2>/dev/null || echo 0); (( cur > size )) && rm -f "$path" |
|
printf '%s/%s/resolve/%s/%s\n' "$HF_ENDPOINT" "$DOWNLOAD_API_PATH" "$REVISION" "$path" |
|
fi |
|
done < "$needed" > "$fileslist_file" |
|
rm -f "$needed" |
|
} |
|
|
|
if should_regenerate_filelist; then |
|
command -v jq &>/dev/null || printf "%b[Warning] jq not installed, using grep/awk for json parsing (slower). Consider installing jq.%b\n" "$YELLOW" "$NC" |
|
|
|
# Total size up front (before the long walk); label the call so its few-second wait isn't silent. |
|
printf "%bFetching repository info...%b" "$DIM" "$NC" |
|
REPO_SIZE=$(fetch_treesize); printf '\r\033[K' |
|
[[ "$REPO_SIZE" =~ ^[0-9]+$ ]] && printf "%bRepository size: %s%b\n" "$DIM" "$(human "$REPO_SIZE")" "$NC" |
|
INCLUDE_REGEX=$(patterns_to_regex "${INCLUDE_PATTERNS[@]}") |
|
EXCLUDE_REGEX=$(patterns_to_regex "${EXCLUDE_PATTERNS[@]}") |
|
printf "%bListing files...%b" "$DIM" "$NC" |
|
# Fast path: typical repos return their whole file list in the metadata; only big repos |
|
# (truncated siblings) need the paginated, resumable tree walk. |
|
if siblings_complete; then |
|
rm -f "$LOCAL_DIR/.hfd/list_state" |
|
emit_siblings | filter_size_path > "$LOCAL_DIR/.hfd/manifest.partial" |
|
gen_status=${PIPESTATUS[0]} |
|
else |
|
walk_tree; gen_status=$? |
|
fi |
|
# A failed walk keeps its checkpoint (list_state + manifest.partial), so a re-run resumes. |
|
(( gen_status == 0 )) || { printf "\n%b[Error] Failed to list repository files. Re-run to resume.%b\n" "$RED" "$NC" >&2; exit 1; } |
|
finalize_filelist |
|
read -r TOTAL_FILES SUM_SIZE < "$LOCAL_DIR/.hfd/filelist_stats" |
|
printf '\r\033[K%bListed %d files (%s)%b\n' "$DIM" "$TOTAL_FILES" "$(human "$SUM_SIZE")" "$NC" |
|
# Whole-revision total from treesize; the summed filtered size when include/exclude narrows it. |
|
TOTAL_SIZE=$REPO_SIZE |
|
[[ -n "$INCLUDE_REGEX$EXCLUDE_REGEX" || ! "$REPO_SIZE" =~ ^[0-9]+$ ]] && TOTAL_SIZE=$SUM_SIZE |
|
printf '%s %s\n' "$TOTAL_FILES" "$TOTAL_SIZE" > "$LOCAL_DIR/.hfd/repo_info" |
|
# Fingerprint written last: marks the list complete (an interrupted walk has none, so it re-lists). |
|
generate_command_string > "$LOCAL_DIR/.hfd/last_download_command" |
|
rm -f "$LOCAL_DIR/.hfd/manifest.partial" "$LOCAL_DIR/.hfd/list_state" "$LOCAL_DIR/.hfd/tree_page.json" |
|
fi |
|
|
|
cd "$LOCAL_DIR" || exit 1 |
|
|
|
# Render one in-place status line; speed = byte delta since last call (PREV_B/PREV_T). |
|
# Fields are fixed-width so columns don't jitter as values grow a digit; ETA is last. |
|
PREV_B=0; PREV_T=0 |
|
render_progress() { |
|
local now=$1 dfiles=$2 dt sp pct eta e |
|
dt=$(( SECONDS - PREV_T )); ((dt<1)) && dt=1 |
|
sp=$(( (now - PREV_B) / dt )); ((sp<0)) && sp=0; PREV_B=$now; PREV_T=$SECONDS |
|
pct=0; ((TOTAL_SIZE>0)) && pct=$(( now * 100 / TOTAL_SIZE )); ((pct>100)) && pct=100 |
|
# ETA only once speed is meaningful; clamp absurd early estimates to keep it tidy. |
|
eta="--:--"; ((sp>0 && TOTAL_SIZE>now)) && { e=$(( (TOTAL_SIZE-now)/sp )); ((e<360000)) && eta=$(printf '%02d:%02d' $((e/60)) $((e%60))); } |
|
printf '\r\033[K%b[%3d%%]%b %*d/%d files | %9s/%9s | %9s/s | ETA %s' \ |
|
"$GREEN" "$pct" "$NC" "${#TOTAL_FILES}" "$dfiles" "$TOTAL_FILES" \ |
|
"$(human "$now")" "$(human "$TOTAL_SIZE")" "$(human "$sp")" "$eta" |
|
} |
|
|
|
# Progress without scanning finished files: manifest totals + a stat of only in-flight files. |
|
monitor_progress() { |
|
local interval=$1 base=$1 stop_hint="Stopping download, please wait..."; PREV_T=$SECONDS |
|
# The main INT trap is deferred until the foreground download returns, so acknowledge |
|
# Ctrl+C here (fires at once in the subshell) and note the force-stop that already works. |
|
[[ "$TOOL" == "aria2c" ]] && stop_hint="Stopping download, please wait... (press Ctrl+C again to force stop)" |
|
trap 'printf "\n%b%s%b\n" "$YELLOW" "$stop_hint" "$NC"; exit 0' INT |
|
if [[ "$TOOL" == "wget" ]]; then |
|
# wget downloads in manifest order, so a forward cursor needs to stat only the |
|
# current file; everything before it is done and counted by its known size. |
|
local -a MS MP; local s p |
|
while IFS=$'\t' read -r s p; do MS+=("$s"); MP+=("$p"); done < .hfd/manifest |
|
local idx=0 done_b=0 cur |
|
while :; do |
|
sleep "$interval" |
|
while (( idx < ${#MP[@]} )); do |
|
cur=$(stat -c%s "${MP[idx]}" 2>/dev/null) || break |
|
(( cur >= MS[idx] )) || break |
|
done_b=$(( done_b + MS[idx] )); idx=$((idx+1)) |
|
done |
|
cur=0; (( idx < ${#MP[@]} )) && cur=$(stat -c%s "${MP[idx]}" 2>/dev/null || echo 0) |
|
render_progress $(( done_b + cur )) "$idx" |
|
done |
|
else |
|
# Parallel downloads finish out of order, so a cursor would miss files completing between |
|
# ticks. Each tick sums block usage (sparse-aware) of the manifest's files only — ignoring |
|
# stale files left from an earlier version of the repo — minus in-progress (.aria2) ones. |
|
local now files t0 walk |
|
while :; do |
|
sleep "$interval" |
|
t0=$SECONDS |
|
read -r now files < <(awk -F'\t' ' |
|
FNR==NR { want[$2]=1; next } |
|
{ p=$2; sub(/^\.\//,"",p) |
|
if (p ~ /\.aria2$/) { sub(/\.aria2$/,"",p); if (p in want) a++; next } |
|
if (p in want) { b+=$1; d++ } } |
|
END { print b*512, d-a } |
|
' .hfd/manifest <(find . -type f ! -path './.hfd/*' -printf '%b\t%p\n' 2>/dev/null)) |
|
render_progress "$now" "$files" |
|
# Self-throttle: if the walk took longer than the base interval, sleep that long |
|
# next time so scanning stays under ~half the wall-clock at any repo size. |
|
walk=$(( SECONDS - t0 )); interval=$base; (( walk > base )) && interval=$walk |
|
done |
|
fi |
|
} |
|
|
|
# Totals were persisted at file-list generation; reload for resumed runs. |
|
[[ -f .hfd/repo_info ]] && read -r TOTAL_FILES TOTAL_SIZE < .hfd/repo_info |
|
TOTAL_FILES=${TOTAL_FILES:-0}; TOTAL_SIZE=${TOTAL_SIZE:-0} |
|
FILE_NOUN="files"; ((TOTAL_FILES==1)) && FILE_NOUN="file" |
|
|
|
# Diff the local tree against the manifest to find what's missing, then short-circuit if done. |
|
((TOTAL_FILES>50000)) && printf "%bChecking local files...%b\n" "$DIM" "$NC" |
|
build_download_list |
|
if (( NEED_COUNT == 0 )); then |
|
printf "%bUp to date. %s %s in %s%b\n" "$GREEN" "$TOTAL_FILES" "$FILE_NOUN" "$LOCAL_DIR" "$NC" |
|
exit 0 |
|
fi |
|
|
|
# "Resuming" if any data is already on disk (completed files or a partial), else a fresh start. |
|
verb="Downloading"; [[ -n "$(find . -type f ! -path './.hfd/*' -print -quit 2>/dev/null)" ]] && verb="Resuming" |
|
printf "%s %s %s to %s · Ctrl+C to stop, re-run to resume\n" "$verb" "$TOTAL_FILES" "$FILE_NOUN" "$LOCAL_DIR" |
|
|
|
# Silence native per-file output (logged to .hfd/download.log) so the monitor owns one |
|
# clean line. Refresh every 1s, backing off for huge repos where the walk gets expensive. |
|
interval=1; ((TOTAL_FILES>50000)) && interval=5 |
|
monitor_progress "$interval" & |
|
MON_PID=$! |
|
trap 'kill "$MON_PID" 2>/dev/null' EXIT |
|
|
|
if [[ "$TOOL" == "aria2c" ]]; then |
|
aria2c --quiet=true --log=.hfd/download.log --log-level=error --file-allocation=none \ |
|
-x "$THREADS" -j "$CONCURRENT" -s "$THREADS" -k 1M -c -i "$fileslist_file" >/dev/null |
|
status=$? |
|
else |
|
wget -x -nH --cut-dirs="$CUT_DIRS" ${HF_TOKEN:+--header="Authorization: Bearer $HF_TOKEN"} \ |
|
--input-file="$fileslist_file" --continue -nv -o .hfd/download.log |
|
status=$? |
|
fi |
|
|
|
# Clear the live progress line in place; the final status line takes its spot (no blank line). |
|
kill "$MON_PID" 2>/dev/null; wait "$MON_PID" 2>/dev/null; printf '\r\033[K' |
|
|
|
if [[ $status -eq 0 ]]; then |
|
printf "%bDone. %s %s, %s in %s%b\n" "$GREEN" "$TOTAL_FILES" "$FILE_NOUN" "$(human "$TOTAL_SIZE")" "$LOCAL_DIR" "$NC" |
|
else |
|
printf "%bDownload incomplete. Re-run to resume. Log: %s%b\n" "$RED" "$PWD/.hfd/download.log" "$NC" |
|
exit 1 |
|
fi |
@padeoe 您好，目前调用脚本无法下载模型，目标路径下只有一个
.hfd文件夹，无模型文件，运行信息如下：已尝试如下方法，均无效：
.hfd文件夹请问应该如何处理？谢谢！