ComfyUI/entrypoint.sh
clsferguson f6d49f33b7
entrypoint: derive correct arch list; add user-tunable build parallelism; fix Sage flags; first-run installs
- Auto-derive TORCH_CUDA_ARCH_LIST from torch device capabilities (unique, sorted, optional +PTX) to cover all charted GPUs:
  Turing 7.5, Ampere 8.0/8.6/8.7, Ada 8.9, Hopper 9.0, and Blackwell 10.0 & 12.0/12.1; add name-based fallbacks for mixed or torch-less scenarios.
- Add user-tunable build parallelism with SAGE_MAX_JOBS (preferred) and MAX_JOBS (alias) that cap PyTorch cpp_extension/ninja -j; fall back to a RAM/CPU heuristic to prevent OOM “Killed” during CUDA/C++ builds.
- Correct Sage flags: SAGE_ATTENTION_AVAILABLE only signals “built/installed,” while FORCE_SAGE_ATTENTION=1 enables Sage at startup; fix logs to reference FORCE_SAGE_ATTENTION.
- Maintain Triton install strategy by GPU generation for compatibility and performance.
- Add first-run dependency installation with COMFY_FORCE_INSTALL override; keep permissions bootstrap and minor logging/URL cleanups.
2025-09-26 22:37:24 -06:00

324 lines
15 KiB
Bash

#!/bin/bash
set -euo pipefail
# --- config ---
APP_USER=${APP_USER:-appuser}
APP_GROUP=${APP_GROUP:-appuser}
PUID=${PUID:-1000}
PGID=${PGID:-1000}
BASE_DIR=/app/ComfyUI
CUSTOM_NODES_DIR="$BASE_DIR/custom_nodes"
SAGE_ATTENTION_DIR="$BASE_DIR/.sage_attention"
SAGE_ATTENTION_BUILT_FLAG="$SAGE_ATTENTION_DIR/.built"
PERMISSIONS_SET_FLAG="$BASE_DIR/.permissions_set"
FIRST_RUN_FLAG="$BASE_DIR/.first_run_done"
# Sage flags:
# - SAGE_ATTENTION_AVAILABLE: set by script to indicate built/importable
# - FORCE_SAGE_ATTENTION=1: force app to enable at startup (default 0)
# --- logging ---
log() { echo "[$(date '+%H:%M:%S')] $1"; }
# --- build parallelism (single knob) ---
# Public knob: SAGE_MAX_JOBS. If unset, pick RAM/CPU heuristic.
decide_build_jobs() {
if [ -n "${SAGE_MAX_JOBS:-}" ]; then echo "$SAGE_MAX_JOBS"; return; fi
local mem_kb=$(awk '/MemTotal:/ {print $2}' /proc/meminfo 2>/dev/null || echo 0)
local cpu=$(nproc) cap=24 jobs
if [ "$mem_kb" -le $((8*1024*1024)) ]; then jobs=2
elif [ "$mem_kb" -le $((12*1024*1024)) ]; then jobs=3
elif [ "$mem_kb" -le $((24*1024*1024)) ]; then jobs=4
elif [ "$mem_kb" -le $((64*1024*1024)) ]; then jobs=$(( cpu<8 ? cpu : 8 ))
else jobs=$cpu; [ "$jobs" -gt "$cap" ] && jobs=$cap
fi
echo "$jobs"
}
# --- CUDA/Torch checks ---
test_pytorch_cuda() {
python -c "
import torch, sys
if not torch.cuda.is_available():
print('[ERROR] PyTorch CUDA not available'); sys.exit(1)
c=torch.cuda.device_count(); print(f'[TEST] PyTorch CUDA available with {c} devices')
for i in range(c):
p=torch.cuda.get_device_properties(i)
print(f'[TEST] GPU {i}: {p.name} (Compute {p.major}.{p.minor})')
" 2>/dev/null
}
# Derive arch list directly from Torch; optional +PTX via SAGE_PTX_FALLBACK=1
compute_arch_list_from_torch() {
python - <<'PY' 2>/dev/null
import os, sys
try:
import torch
if not torch.cuda.is_available():
print(""); sys.exit(0)
caps = {f"{torch.cuda.get_device_properties(i).major}.{torch.cuda.get_device_properties(i).minor}"
for i in range(torch.cuda.device_count())}
ordered = sorted(caps, key=lambda s: tuple(int(x) for x in s.split(".")))
if not ordered: print(""); sys.exit(0)
if os.environ.get("SAGE_PTX_FALLBACK","0")=="1":
highest = ordered[-1]; print(";".join(ordered+[highest + "+PTX"]))
else:
print(";".join(ordered))
except Exception:
print("")
PY
}
# Fallback name-based mapping across Turing→Blackwell
detect_gpu_generations() {
local info=$(nvidia-smi --query-gpu=name --format=csv,noheader,nounits 2>/dev/null || echo "")
local has_turing=false has_amp_ga100=false has_amp_ga10x=false has_amp_ga10b=false
local has_ada=false has_hopper=false has_bw_cons=false has_bw_dc=false
local n=0
[ -z "$info" ] && { log "No NVIDIA GPUs detected"; return 1; }
log "Detecting GPU generations:"
while IFS= read -r g; do
n=$((n+1)); log " GPU $n: $g"
case "$g" in
*"RTX 20"*|*"T4"*) has_turing=true ;;
*"A100"*|*"A30"*|*"A40"*) has_amp_ga100=true ;;
*"RTX 30"*|*"RTX 3090"*|*"RTX 3080"*|*"RTX 3070"*|*"RTX 3060"*) has_amp_ga10x=true ;;
*"Orin"*|*"Jetson"*) has_amp_ga10b=true ;;
*"RTX 40"*|*"4090"*|*"L40"*|*"L4"*) has_ada=true ;;
*"H100"*|*"H200"*|*"GH200"*) has_hopper=true ;;
*"RTX 50"*|*"5090"*|*"5080"*|*"5070"*|*"5060"*|*"PRO "*Blackwell*|*"PRO 4000 Blackwell"*) has_bw_cons=true ;;
*"B200"*|*"B100"*|*"GB200"*|*"B40"*|*"RTX 6000 Blackwell"*|*"RTX 5000 Blackwell"*) has_bw_dc=true ;;
esac
done <<< "$info"
export DET_TURING=$has_turing DET_AMP80=$has_amp_ga100 DET_AMP86=$has_amp_ga10x DET_AMP87=$has_amp_ga10b
export DET_ADA=$has_ada DET_HOPPER=$has_hopper DET_BW12=$has_bw_cons DET_BW10=$has_bw_dc
export GPU_COUNT=$n
log "Summary: Turing=$has_turing Amp(8.0)=$has_amp_ga100 Amp(8.6)=$has_amp_ga10x Amp(8.7)=$has_amp_ga10b Ada=$has_ada Hopper=$has_hopper Blackwell(12.x)=$has_bw_cons Blackwell(10.0)=$has_bw_dc"
test_pytorch_cuda && log "PyTorch CUDA compatibility confirmed" || log "WARNING: PyTorch CUDA compatibility issues detected"
}
determine_sage_strategy() {
local s=""
if [ "${DET_TURING:-false}" = "true" ]; then
if [ "${DET_AMP80:-false}" = "true" ] || [ "${DET_AMP86:-false}" = "true" ] || [ "${DET_AMP87:-false}" = "true" ] || [ "${DET_ADA:-false}" = "true" ] || [ "${DET_HOPPER:-false}" = "true" ] || [ "${DET_BW12:-false}" = "true" ] || [ "${DET_BW10:-false}" = "true" ]; then
s="mixed_with_turing"; log "Mixed rig including Turing - using compatibility mode"
else s="turing_only"; log "Turing-only rig detected"; fi
elif [ "${DET_BW12:-false}" = "true" ] || [ "${DET_BW10:-false}" = "true" ]; then s="blackwell_capable"; log "Blackwell detected - using latest optimizations"
elif [ "${DET_HOPPER:-false}" = "true" ]; then s="hopper_capable"; log "Hopper detected - using modern optimizations"
elif [ "${DET_ADA:-false}" = "true" ] || [ "${DET_AMP86:-false}" = "true" ] || [ "${DET_AMP87:-false}" = "true" ] || [ "${DET_AMP80:-false}" = "true" ]; then
s="ampere_ada_optimized"; log "Ampere/Ada detected - using standard optimizations"
else s="fallback"; log "Unknown configuration - using fallback"; fi
export SAGE_STRATEGY=$s
}
install_triton_version() {
case "$SAGE_STRATEGY" in
"mixed_with_turing"|"turing_only")
log "Installing Triton 3.2.0 for Turing compatibility"
python -m pip install --user --force-reinstall "triton==3.2.0" || python -m pip install --user --force-reinstall triton || true
;;
"blackwell_capable"|"hopper_capable")
log "Installing latest Triton for Hopper/Blackwell"
python -m pip install --user --force-reinstall triton || python -m pip install --user --force-reinstall --pre triton || python -m pip install --user --force-reinstall "triton>=3.2.0" || true
;;
*)
log "Installing latest stable Triton"
python -m pip install --user --force-reinstall triton || { log "WARNING: Triton installation failed"; return 1; }
;;
esac
}
build_sage_attention_mixed() {
log "Building Sage Attention..."
mkdir -p "$SAGE_ATTENTION_DIR"; cd "$SAGE_ATTENTION_DIR"
local arch_list="${SAGE_ARCH_LIST_OVERRIDE:-$(compute_arch_list_from_torch)}"
if [ -z "$arch_list" ]; then
local tmp=""
[ "${DET_TURING:-false}" = "true" ] && tmp="${tmp}7.5;"
[ "${DET_AMP80:-false}" = "true" ] && tmp="${tmp}8.0;"
[ "${DET_AMP86:-false}" = "true" ] && tmp="${tmp}8.6;"
[ "${DET_AMP87:-false}" = "true" ] && tmp="${tmp}8.7;"
[ "${DET_ADA:-false}" = "true" ] && tmp="${tmp}8.9;"
[ "${DET_HOPPER:-false}" = "true" ] && tmp="${tmp}9.0;"
[ "${DET_BW10:-false}" = "true" ] && tmp="${tmp}10.0;"
[ "${DET_BW12:-false}" = "true" ] && tmp="${tmp}12.0;"
arch_list="${tmp%;}"
fi
export TORCH_CUDA_ARCH_LIST="$arch_list"
log "Set TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST"
case "$SAGE_STRATEGY" in
"mixed_with_turing"|"turing_only")
log "Cloning SageAttention v1.0 for Turing"
if [ -d "SageAttention/.git" ]; then cd SageAttention; git fetch --depth 1 origin || return 1; git checkout v1.0 2>/dev/null || git checkout -b v1.0 origin/v1.0 || return 1; git reset --hard origin/v1.0 || return 1
else rm -rf SageAttention; git clone --depth 1 https://github.com/thu-ml/SageAttention.git -b v1.0 || return 1; cd SageAttention; fi
;;
*)
log "Cloning latest SageAttention"
if [ -d "SageAttention/.git" ]; then cd SageAttention; git fetch --depth 1 origin || return 1; git reset --hard origin/main || return 1
else rm -rf SageAttention; git clone --depth 1 https://github.com/thu-ml/SageAttention.git || return 1; cd SageAttention; fi
;;
esac
[ "${SAGE_VERBOSE_BUILD:-0}" = "1" ] && export TORCH_CPP_BUILD_VERBOSE=1
local jobs; jobs="$(decide_build_jobs)"
log "Using MAX_JOBS=${jobs} for SageAttention build"
# Set MAX_JOBS only for this build call to avoid leaking globally
if MAX_JOBS="${jobs}" python -m pip install --user --no-build-isolation .; then
echo "$SAGE_STRATEGY|$TORCH_CUDA_ARCH_LIST" > "$SAGE_ATTENTION_BUILT_FLAG"
log "SageAttention built successfully"
cd "$BASE_DIR"; return 0
else
log "ERROR: SageAttention build failed"
cd "$BASE_DIR"; return 1
fi
}
needs_rebuild() {
if [ ! -f "$SAGE_ATTENTION_BUILT_FLAG" ]; then return 0; fi
local x; x=$(cat "$SAGE_ATTENTION_BUILT_FLAG" 2>/dev/null || echo "")
local prev_strategy="${x%%|*}"; local prev_arch="${x#*|}"
if [ "$prev_strategy" != "$SAGE_STRATEGY" ] || [ "$prev_arch" != "$TORCH_CUDA_ARCH_LIST" ]; then return 0; fi
return 1
}
test_sage_attention() {
python -c "
import sys
try:
import sageattention; print('[TEST] SageAttention import: SUCCESS')
v=getattr(sageattention,'__version__',None)
if v: print(f'[TEST] Version: {v}'); sys.exit(0)
except ImportError as e:
print(f'[TEST] SageAttention import: FAILED - {e}'); sys.exit(1)
except Exception as e:
print(f'[TEST] SageAttention test: ERROR - {e}'); sys.exit(1)
" 2>/dev/null
}
setup_sage_attention() {
export SAGE_ATTENTION_BUILT=0 SAGE_ATTENTION_AVAILABLE=0
if ! detect_gpu_generations; then log "No GPUs detected, skipping SageAttention setup"; return 0; fi
determine_sage_strategy
# Resolve arch list early
export TORCH_CUDA_ARCH_LIST="${SAGE_ARCH_LIST_OVERRIDE:-$(compute_arch_list_from_torch)}"
if [ -z "$TORCH_CUDA_ARCH_LIST" ]; then
local tmp=""
[ "${DET_TURING:-false}" = "true" ] && tmp="${tmp}7.5;"
[ "${DET_AMP80:-false}" = "true" ] && tmp="${tmp}8.0;"
[ "${DET_AMP86:-false}" = "true" ] && tmp="${tmp}8.6;"
[ "${DET_AMP87:-false}" = "true" ] && tmp="${tmp}8.7;"
[ "${DET_ADA:-false}" = "true" ] && tmp="${tmp}8.9;"
[ "${DET_HOPPER:-false}" = "true" ] && tmp="${tmp}9.0;"
[ "${DET_BW10:-false}" = "true" ] && tmp="${tmp}10.0;"
[ "${DET_BW12:-false}" = "true" ] && tmp="${tmp}12.0;"
export TORCH_CUDA_ARCH_LIST="${tmp%;}"
fi
log "Resolved TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST"
if needs_rebuild || ! test_sage_attention; then
log "Building SageAttention..."
if install_triton_version && build_sage_attention_mixed && test_sage_attention; then
export SAGE_ATTENTION_BUILT=1 SAGE_ATTENTION_AVAILABLE=1
log "SageAttention is built; set FORCE_SAGE_ATTENTION=1 to enable it at startup"
else
export SAGE_ATTENTION_BUILT=0 SAGE_ATTENTION_AVAILABLE=0
log "WARNING: SageAttention is not available after build attempt"
fi
else
export SAGE_ATTENTION_BUILT=1 SAGE_ATTENTION_AVAILABLE=1
log "SageAttention already built and importable"
fi
}
# --- root to runtime user ---
if [ "$(id -u)" = "0" ]; then
if [ ! -f "$PERMISSIONS_SET_FLAG" ]; then
log "Setting up user permissions..."
if getent group "${PGID}" >/dev/null; then
EXISTING_GRP="$(getent group "${PGID}" | cut -d: -f1)"; usermod -g "${EXISTING_GRP}" "${APP_USER}" || true; APP_GROUP="${EXISTING_GRP}"
else groupmod -o -g "${PGID}" "${APP_GROUP}" || true; fi
usermod -o -u "${PUID}" "${APP_USER}" || true
mkdir -p "/home/${APP_USER}"
for d in "$BASE_DIR" "/home/$APP_USER"; do [ -e "$d" ] && chown -R "${APP_USER}:${APP_GROUP}" "$d" || true; done
readarray -t PY_PATHS < <(python - <<'PY'
import sys, sysconfig, os, datetime
def log(m): print(f"[bootstrap:python {datetime.datetime.now().strftime('%H:%M:%S')}] {m}", file=sys.stderr, flush=True)
log("Determining writable Python install targets via sysconfig.get_paths()")
for k in ("purelib","platlib","scripts","include","platinclude","data"):
v = sysconfig.get_paths().get(k)
if v: print(v); log(f"emit {k} -> {v}")
d = sysconfig.get_paths().get("data")
if d:
share=os.path.join(d,"share"); man1=os.path.join(share,"man","man1")
print(share); print(man1); log(f"emit wheel data dirs -> {share}, {man1}")
PY
)
for d in "${PY_PATHS[@]}"; do
case "$d" in
/usr/local|/usr/local/*) mkdir -p "$d" || true; chown -R "${APP_USER}:${APP_GROUP}" "$d" || true; chmod -R u+rwX,g+rwX "$d" || true ;;
*) : ;;
esac
done
touch "$PERMISSIONS_SET_FLAG"; chown "${APP_USER}:${APP_GROUP}" "$PERMISSIONS_SET_FLAG"
log "User permissions configured"
else log "User permissions already configured, skipping..."; fi
exec runuser -u "${APP_USER}" -- "$0" "$@"
fi
# --- SageAttention setup ---
setup_sage_attention
# --- ComfyUI-Manager sync ---
if [ -d "$CUSTOM_NODES_DIR/ComfyUI-Manager/.git" ]; then
log "Updating ComfyUI-Manager"; git -C "$CUSTOM_NODES_DIR/ComfyUI-Manager" fetch --depth 1 origin || true
git -C "$CUSTOM_NODES_DIR/ComfyUI-Manager" reset --hard origin/HEAD || true; git -C "$CUSTOM_NODES_DIR/ComfyUI-Manager" clean -fdx || true
elif [ ! -d "$CUSTOM_NODES_DIR/ComfyUI-Manager" ]; then
log "Installing ComfyUI-Manager"; git clone --depth 1 https://github.com/ltdrdata/ComfyUI-Manager.git "$CUSTOM_NODES_DIR/ComfyUI-Manager" || true
fi
# --- PATH/PYTHONPATH ---
export PATH="$HOME/.local/bin:$PATH"
pyver="$(python -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")')"
export PYTHONPATH="$HOME/.local/lib/python${pyver}/site-packages:${PYTHONPATH:-}"
# --- first-run install of custom_nodes ---
if [ ! -f "$FIRST_RUN_FLAG" ] || [ "${COMFY_FORCE_INSTALL:-0}" = "1" ]; then
if [ "${COMFY_AUTO_INSTALL:-1}" = "1" ]; then
log "First run or forced; installing custom node dependencies..."
while IFS= read -r -d '' req; do
log "python -m pip install --user --upgrade -r $req"
python -m pip install --no-cache-dir --user --upgrade --upgrade-strategy only-if-needed -r "$req" || true
done < <(find "$CUSTOM_NODES_DIR" -maxdepth 3 -type f \( -iname 'requirements.txt' -o -iname 'requirements-*.txt' -o -path '*/requirements/*.txt' \) -print0)
while IFS= read -r -d '' pjt; do
d="$(dirname "$pjt")"; log "python -m pip install --user . in $d"
(cd "$d" && python -m pip install --no-cache-dir --user .) || true
done < <(find "$CUSTOM_NODES_DIR" -maxdepth 2 -type f -iname 'pyproject.toml' -not -path '*/ComfyUI-Manager/*' -print0)
python -m pip check || true
else log "COMFY_AUTO_INSTALL=0; skipping dependency install"; fi
touch "$FIRST_RUN_FLAG"
else
log "Not first run; skipping custom_nodes dependency install"
fi
# --- launch ComfyUI ---
COMFYUI_ARGS=""
if [ "${FORCE_SAGE_ATTENTION:-0}" = "1" ]; then
if test_sage_attention; then COMFYUI_ARGS="--use-sage-attention"; log "Starting ComfyUI with SageAttention (FORCE_SAGE_ATTENTION=1)"
else log "WARNING: FORCE_SAGE_ATTENTION=1 but import failed; starting without"; fi
else
if [ "${SAGE_ATTENTION_AVAILABLE:-0}" = "1" ]; then log "SageAttention is built; set FORCE_SAGE_ATTENTION=1 to enable"
else log "SageAttention not available; starting without it"; fi
fi
cd "$BASE_DIR"
if [ $# -eq 0 ]; then exec python main.py --listen 0.0.0.0 $COMFYUI_ARGS
else
if [ "$1" = "python" ] && [ "${2:-}" = "main.py" ]; then shift 2; exec python main.py $COMFYUI_ARGS "$@"
else exec "$@"; fi
fi