Repository navigation
Expand file tree
/
Copy pathdeploy.sh
More file actions
executable file
·477 lines (444 loc) · 18.7 KB
/
Copy pathdeploy.sh
File metadata and controls
executable file
·477 lines (444 loc) · 18.7 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
#!/bin/bash
# deploy.sh — ComfyCLI 纯二进制部署打包脚本
#
# 把编译好的二进制 + 所有依赖 .so 打包,远程无需 Python/pip/venv。
# 用法:
# ./deploy.sh # 打包部署包到 dist/
# ./deploy.sh --scp user@host # 打包 + SCP 到远程
# GLIBC_TARGET=2.35 ./deploy.sh --scp user@host # 指定目标 GLIBC 版本
# WITH_CUDA=1 ./deploy.sh --scp user@host # 同时打包 CUDA Runtime
#
# 前提:先用 build.sh 编译好 comfycli-bin + comfycli-bin.so + libsdcpp_adapter.so
#
# 设计原则:
# - 不编译,只打包部署(编译用 build.sh)
# - GLIBC 兼容:从 /opt/deb/<version>/ 取目标版本 GLIBC + libstdc++
# 打包到 dist/lib/,远程通过 run.sh 设 LD_LIBRARY_PATH 优先加载
# - 可选打包 CUDA Runtime(libcudart/cublas/cublasLt)到 dist/lib/
# - run.sh 自动设置 LD_LIBRARY_PATH,远程一条命令启动
#
# 准备 deb(一次性的):
# ssh root@remote_host "ldd --version | head -1" # 看远程 GLIBC 版本
# mkdir -p /opt/deb/2.35 && cd /opt/deb/2.35
# wget http://archive.ubuntu.com/ubuntu/pool/main/g/glibc/libc6_2.35-0ubuntu3_amd64.deb
# wget http://archive.ubuntu.com/ubuntu/pool/main/g/glibc/libc6-dev_2.35-0ubuntu3_amd64.deb
# wget http://archive.ubuntu.com/ubuntu/pool/main/g/gcc-11/libstdc++-11-dev_11.4.0-1ubuntu1~22.04_amd64.deb
# for d in *.deb; do dpkg-deb -x "$d" . ; done
set -euo pipefail
PROJECT_DIR="$(cd "$(dirname "$0")" && pwd)"
DIST_DIR="$PROJECT_DIR/dist"
SD_BACKEND_DL="${SD_BACKEND_DL:-1}"
SD_BUILD_DIR="${SD_BUILD_DIR:-/opt/sd/build-dl}"
WITH_CUDA_BACKEND="${WITH_CUDA_BACKEND:-1}"
WITH_CUDA="${WITH_CUDA:-0}"
CUDA_DIR="${CUDA_DIR:-/data/cuda/targets/x86_64-linux/lib}"
# Florence-2 图像→提示词节点(libflorence2.so + ONNX Runtime + OpenCV 依赖闭包)。
# 默认打包(远程用 CPU 跑 Florence,避免 344MB 的 CUDA provider);
# WITH_FLORENCE2_MODELS=1 时附带 ~1.1GB 的 ONNX 权重目录。
WITH_FLORENCE2="${WITH_FLORENCE2:-1}"
WITH_FLORENCE2_MODELS="${WITH_FLORENCE2_MODELS:-0}"
# WITH_FLORENCE2_CUDA:打包 ONNX Runtime CUDA provider(Florence 走 GPU,+329MB)。
# 默认 1 —— 常规远程环境都自带 CUDA runtime 与 cuDNN 9,直接用 GPU;
# 完全全新的环境用 WITH_FLORENCE2_CUDA=0 退回 CPU(或 =1 时缺依赖会自动回落 CPU)。
# WITH_FLORENCE2_CUDA_FULL=1 连 CUDA runtime(cudart/cublas/cublasLt/curand/cufft)
# 一起打包(+~0.9GB;cuDNN 仍不打包,太大)。
WITH_FLORENCE2_CUDA="${WITH_FLORENCE2_CUDA:-1}"
WITH_FLORENCE2_CUDA_FULL="${WITH_FLORENCE2_CUDA_FULL:-0}"
FLORENCE2_MODEL_DIR="${FLORENCE2_MODEL_DIR:-/data/models/florence2}"
# BiRefNet 抠图节点(libbirefnet.so,复用同一份 ONNX Runtime)。
WITH_BIREFNET="${WITH_BIREFNET:-1}"
WITH_BIREFNET_MODELS="${WITH_BIREFNET_MODELS:-0}"
BIREFNET_MODEL_DIR="${BIREFNET_MODEL_DIR:-/data/models/birefnet}"
# OCR 节点(libocr.so,PP-OCRv4;模型仅 ~15MB)。
WITH_OCR="${WITH_OCR:-1}"
WITH_OCR_MODELS="${WITH_OCR_MODELS:-1}"
OCR_MODEL_DIR="${OCR_MODEL_DIR:-/data/models/ocr}"
# ControlNet 条件图预处理器(preprocesscli,复用 ONNX Runtime + OpenCV)。
WITH_PREPROCESS="${WITH_PREPROCESS:-1}"
WITH_PREPROCESS_MODELS="${WITH_PREPROCESS_MODELS:-0}"
PREPROCESS_MODEL_DIR="${PREPROCESS_MODEL_DIR:-/data/models/image}"
GLIBC_TARGET="${GLIBC_TARGET:-}"
if [ -n "$GLIBC_TARGET" ]; then
GLIBC_SYSROOT="/opt/deb/$GLIBC_TARGET"
TARBALL_SUFFIX="_glibc${GLIBC_TARGET}"
else
GLIBC_SYSROOT=""
TARBALL_SUFFIX=""
fi
if [ "$WITH_CUDA" = "1" ]; then
TARBALL_SUFFIX="${TARBALL_SUFFIX}_cuda"
WITH_CUDA_BACKEND=1
fi
if [ "$WITH_CUDA_BACKEND" = "1" ] && [ ! -f "$SD_BUILD_DIR/bin/libggml-cuda.so" ]; then
echo "警告: WITH_CUDA_BACKEND=1 但找不到 $SD_BUILD_DIR/bin/libggml-cuda.so,已回退为 CPU-only"
WITH_CUDA_BACKEND=0
fi
echo "============================================"
echo " ComfyCLI 纯二进制部署包"
if [ "$SD_BACKEND_DL" = "1" ]; then
echo " 后端模式: 动态加载 (libggml-cuda.so 可独立分发)"
else
echo " 后端模式: 静态链接"
fi
if [ "$WITH_CUDA_BACKEND" = "1" ]; then
echo " CUDA 后端插件: 包含"
else
echo " CUDA 后端插件: 不包含(CPU-only)"
fi
if [ "$WITH_CUDA" = "1" ]; then
echo " CUDA Runtime: 打包"
else
echo " CUDA Runtime: 不打包(远程需自带 CUDA Runtime)"
fi
if [ "$WITH_FLORENCE2" = "1" ] && [ -f "$PROJECT_DIR/cpp/florence2/libflorence2.so" ]; then
if [ "$WITH_FLORENCE2_CUDA" = "1" ]; then
echo " Florence-2 节点: 打包 (GPU,含 CUDA provider)"
else
echo " Florence-2 节点: 打包 (CPU 推理)"
fi
[ "$WITH_FLORENCE2_MODELS" = "1" ] && echo " Florence-2 权重: 打包 (~1.1GB)"
else
echo " Florence-2 节点: 不打包"
fi
if [ "$WITH_BIREFNET" = "1" ] && [ -f "$PROJECT_DIR/cpp/birefnet/libbirefnet.so" ]; then
echo " BiRefNet 抠图节点: 打包"
[ "$WITH_BIREFNET_MODELS" = "1" ] && echo " BiRefNet 权重: 打包 (~973MB)"
fi
if [ "$WITH_OCR" = "1" ] && [ -f "$PROJECT_DIR/cpp/ocr/libocr.so" ]; then
echo " OCR 节点: 打包 (PP-OCRv4)"
[ "$WITH_OCR_MODELS" = "1" ] && echo " OCR 模型: 打包 (~15MB)"
fi
if [ "$WITH_PREPROCESS" = "1" ] && [ -f "$PROJECT_DIR/cpp/preprocess/preprocesscli" ]; then
echo " 条件图预处理: 打包 (preprocesscli)"
[ "$WITH_PREPROCESS_MODELS" = "1" ] && echo " 预处理模型: 打包 (~510MB)"
fi
echo "============================================"
# ── 检查二进制是否存在 ──
if [ ! -f "$PROJECT_DIR/comfycli-bin" ] || [ ! -f "$PROJECT_DIR/comfycli-bin.so" ]; then
echo "错误: 找不到 comfycli-bin 或 comfycli-bin.so"
echo "请先运行 build.sh 编译"
exit 1
fi
# ── 清空部署目录 ──
rm -rf "$DIST_DIR"
mkdir -p "$DIST_DIR/lib"
# ── 收集依赖 .so 到 dist/lib/ ──
echo ""
echo ">>> 收集依赖 .so 到 $DIST_DIR/lib"
GLIBC_SRC="/lib/x86_64-linux-gnu"
if [ -n "$GLIBC_TARGET" ]; then
GLIBC_SRC2="$GLIBC_SYSROOT/lib/x86_64-linux-gnu"
[ -d "$GLIBC_SRC2" ] && GLIBC_SRC="$GLIBC_SRC2"
fi
echo " GLIBC 源: $GLIBC_SRC"
LIBMISSING=0
for lib in libm.so.6 libc.so.6 libpthread.so.0 librt.so.1 libdl.so.2 ld-linux-x86-64.so.2 libstdc++.so.6 libgomp.so.1; do
f=$(find "$GLIBC_SRC" -name "$lib" 2>/dev/null | head -1)
if [ -n "$f" ]; then
cp -L "$f" "$DIST_DIR/lib/"
echo " ✓ $lib"
else
echo " ✗ $lib 未找到"
LIBMISSING=$((LIBMISSING + 1))
fi
done
echo ""
echo " 共 $(ls "$DIST_DIR/lib" | wc -l) 个 .so 文件"
if [ "$LIBMISSING" -gt 0 ]; then
echo " 警告: $LIBMISSING 个 GLIBC 兼容库未找到"
fi
# ── 可选:打包 CUDA Runtime ──
if [ "$WITH_CUDA" = "1" ]; then
echo ""
echo ">>> 收集 CUDA Runtime .so 到 $DIST_DIR/lib"
if [ ! -d "$CUDA_DIR" ]; then
echo " 错误: CUDA 目录不存在: $CUDA_DIR"
exit 1
fi
CUDA_MISSING=0
for lib in libcudart.so.12 libcublas.so.12 libcublasLt.so.12; do
f=$(find "$CUDA_DIR" -name "$lib" 2>/dev/null | head -1)
if [ -n "$f" ]; then
cp -L "$f" "$DIST_DIR/lib/"
echo " ✓ $lib"
else
echo " ✗ $lib 未找到"
CUDA_MISSING=$((CUDA_MISSING + 1))
fi
done
if [ "$CUDA_MISSING" -gt 0 ]; then
echo " 警告: $CUDA_MISSING 个 CUDA Runtime 库未找到"
exit 1
fi
fi
# ── 复制二进制到 dist/ ──
echo ""
echo ">>> 复制二进制"
cp "$PROJECT_DIR/comfycli-bin" "$DIST_DIR/"
cp "$PROJECT_DIR/comfycli-bin.so" "$DIST_DIR/"
# 复制 C++ 推理后端
ADAPTER_SO=""
if [ -f "$PROJECT_DIR/cpp/sd/build/libsdcpp_adapter.so" ]; then
ADAPTER_SO="$PROJECT_DIR/cpp/sd/build/libsdcpp_adapter.so"
elif [ -f "$PROJECT_DIR/libsdcpp_adapter.so" ]; then
ADAPTER_SO="$PROJECT_DIR/libsdcpp_adapter.so"
fi
if [ -n "$ADAPTER_SO" ]; then
cp "$ADAPTER_SO" "$DIST_DIR/"
# comfycli-bin 通过 dlopen 加载 libsdcpp_adapter.so,其依赖(opencv→gdal/gdcm 链)
# 不在 comfycli-bin 的 ldd 里,需递归收集并按 SONAME 命名放进 lib/。
echo ""
echo ">>> 收集 sd adapter 依赖闭包 (opencv/gdal/gdcm...)"
python3 - "$ADAPTER_SO" "$DIST_DIR/lib" <<'PY'
import subprocess, os, re, sys, shutil
adapter, dst = sys.argv[1], sys.argv[2]
base = {'libc.so.6','libm.so.6','libpthread.so.0','librt.so.1','libdl.so.2','ld-linux-x86-64.so.2',
'libstdc++.so.6','libgcc_s.so.1','libresolv.so.2','libutil.so.1','libnsl.so.1','libanl.so.1',
'libmvec.so.1','libcuda.so.1'}
def ldd(p):
out = subprocess.run(['ldd', p], capture_output=True, text=True).stdout
d = {}
for line in out.splitlines():
m = re.match(r'\s*(\S+)\s+=>\s+(\S+)', line)
if m and m.group(2).startswith('/'):
d[m.group(1)] = os.path.realpath(m.group(2))
return d
seen, todo = {}, [adapter]
while todo:
p = todo.pop()
for name, path in ldd(p).items():
if name in seen or name in base or '/' in name:
continue
seen[name] = path
todo.append(path)
n = 0
for name, path in seen.items():
b = os.path.basename(path)
if b.startswith(('libstable-diffusion', 'libggml')): # 已由 SD_LIBS 段按版本名复制
continue
if os.path.exists(path):
shutil.copy2(path, os.path.join(dst, name))
n += 1
print(f" ✓ adapter 闭包 {n} 个 .so -> lib/")
PY
else
echo "警告: libsdcpp_adapter.so 未找到"
fi
# ── 可选:Florence-2 图像→提示词(libflorence2.so + ONNX Runtime + OpenCV)──
if [ "$WITH_FLORENCE2" = "1" ] && [ -f "$PROJECT_DIR/cpp/florence2/libflorence2.so" ]; then
echo ""
echo ">>> 打包 Florence-2 节点"
cp "$PROJECT_DIR/cpp/florence2/libflorence2.so" "$DIST_DIR/"
echo " ✓ libflorence2.so"
# 依赖闭包:libflorence2.so + 其 RUNPATH 解析出的 libonnxruntime/opencv + 依赖。
# 排除 GLIBC 核心(lib/ 兼容层已提供)、libcuda(远程驱动提供)、libstdc++/gomp(已复制)。
FLOR2="$PROJECT_DIR/cpp/florence2/libflorence2.so"
LDD_UNION="$(mktemp)"
ldd "$FLOR2" 2>/dev/null | grep -oP '=> \K[^ ]+' | grep '^/' | sort -u >> "$LDD_UNION" || true
# ONNX Runtime 的 provider 共享库由 libonnxruntime dlopen,不在 ldd 里,显式补上
ORT_REAL=$(ldd "$FLOR2" 2>/dev/null | awk '/libonnxruntime\.so\.1 /{print $3}')
if [ -n "$ORT_REAL" ]; then
ORT_DIR="$(dirname "$ORT_REAL")"
for extra in "$ORT_DIR"/libonnxruntime_providers_shared.so; do
[ -f "$extra" ] && echo "$extra" >> "$LDD_UNION"
done
fi
F2_COPIED=0
while read -r f; do
case "$f" in
*/libc.so.6|*/libm.so.6|*/libpthread.so.0|*/libdl.so.2|*/librt.so.1|*/ld-linux-x86-64.so.2|*/libresolv.so.2|*/libgcc_s.so.1|*/libcuda.so.1|*/libstdc++.so.6|*/libgomp.so.1) continue ;;
esac
cp -L "$f" "$DIST_DIR/lib/" && F2_COPIED=$((F2_COPIED + 1))
done < "$LDD_UNION"
rm -f "$LDD_UNION"
echo " ✓ 依赖闭包 $F2_COPIED 个 .so -> lib/"
if [ "$WITH_FLORENCE2_CUDA" = "1" ] && [ -n "$ORT_REAL" ]; then
ORT_DIR="$(dirname "$ORT_REAL")"
if [ -f "$ORT_DIR/libonnxruntime_providers_cuda.so" ]; then
cp -L "$ORT_DIR/libonnxruntime_providers_cuda.so" "$DIST_DIR/lib/"
echo " ✓ libonnxruntime_providers_cuda.so (Florence 走 GPU)"
if [ "$WITH_FLORENCE2_CUDA_FULL" = "1" ]; then
ldd "$ORT_DIR/libonnxruntime_providers_cuda.so" 2>/dev/null | grep -oP '=> \K[^ ]+' | grep '^/' | sort -u | while read -r f; do
case "$f" in
*/libc.so.6|*/libm.so.6|*/libpthread.so.0|*/libdl.so.2|*/librt.so.1|*/ld-linux-x86-64.so.2|*/libresolv.so.2|*/libgcc_s.so.1|*/libcuda.so.1|*/libstdc++.so.6|*/libgomp.so.1|*/libcudnn*) continue ;;
esac
cp -L "$f" "$DIST_DIR/lib/" 2>/dev/null || true
done
echo " ✓ CUDA runtime 依赖已打包 (FULL)"
else
echo " · CUDA runtime/cuDNN 未打包;远程需自带(缺失则自动回落 CPU)"
fi
else
echo " ⚠ 未找到 libonnxruntime_providers_cuda.so,Florence 将走 CPU"
fi
fi
if [ "$WITH_FLORENCE2_MODELS" = "1" ]; then
if [ -d "$FLORENCE2_MODEL_DIR" ]; then
mkdir -p "$DIST_DIR/florence2"
cp -r "$FLORENCE2_MODEL_DIR/onnx" "$DIST_DIR/florence2/" 2>/dev/null || true
[ -f "$FLORENCE2_MODEL_DIR/vocab.bin" ] && cp "$FLORENCE2_MODEL_DIR/vocab.bin" "$DIST_DIR/florence2/"
echo " ✓ Florence-2 权重 -> florence2/"
else
echo " ⚠ 找不到模型目录 $FLORENCE2_MODEL_DIR,跳过权重"
fi
else
echo " · 权重未打包;远程需把模型放到 $FLORENCE2_MODEL_DIR(或用 FLORENCE2_MODEL_DIR 指向)"
fi
fi
# ── 可选:BiRefNet 抠图(libbirefnet.so,复用同一份 ONNX Runtime)──
if [ "$WITH_BIREFNET" = "1" ] && [ -f "$PROJECT_DIR/cpp/birefnet/libbirefnet.so" ]; then
echo ""
echo ">>> 打包 BiRefNet 抠图节点"
cp "$PROJECT_DIR/cpp/birefnet/libbirefnet.so" "$DIST_DIR/"
echo " ✓ libbirefnet.so"
if [ "$WITH_BIREFNET_MODELS" = "1" ] && [ -d "$BIREFNET_MODEL_DIR" ]; then
mkdir -p "$DIST_DIR/birefnet"
cp -r "$BIREFNET_MODEL_DIR/onnx" "$DIST_DIR/birefnet/" 2>/dev/null || true
[ -f "$BIREFNET_MODEL_DIR/preprocessor_config.json" ] && cp "$BIREFNET_MODEL_DIR/preprocessor_config.json" "$DIST_DIR/birefnet/"
echo " ✓ BiRefNet 权重 -> birefnet/"
else
echo " · 权重未打包;远程需放 $BIREFNET_MODEL_DIR(或用 BIREFNET_MODEL 指向)"
fi
fi
# ── 可选:OCR(libocr.so,PP-OCRv4,复用同一份 ONNX Runtime)──
if [ "$WITH_OCR" = "1" ] && [ -f "$PROJECT_DIR/cpp/ocr/libocr.so" ]; then
echo ""
echo ">>> 打包 OCR 节点"
cp "$PROJECT_DIR/cpp/ocr/libocr.so" "$DIST_DIR/"
echo " ✓ libocr.so"
if [ "$WITH_OCR_MODELS" = "1" ] && [ -d "$OCR_MODEL_DIR" ]; then
mkdir -p "$DIST_DIR/ocr"
cp "$OCR_MODEL_DIR"/ch_PP-OCRv4_det_infer.onnx "$OCR_MODEL_DIR"/ch_PP-OCRv4_rec_infer.onnx \
"$OCR_MODEL_DIR"/ppocr_keys_v1.txt "$DIST_DIR/ocr/" 2>/dev/null || true
echo " ✓ PP-OCRv4 模型 -> ocr/"
fi
fi
# ── 可选:ControlNet 条件图预处理器(preprocesscli,复用 ONNX Runtime + OpenCV)──
if [ "$WITH_PREPROCESS" = "1" ] && [ -f "$PROJECT_DIR/cpp/preprocess/preprocesscli" ]; then
echo ""
echo ">>> 打包 ControlNet 条件图预处理器"
cp "$PROJECT_DIR/cpp/preprocess/preprocesscli" "$DIST_DIR/"
chmod +x "$DIST_DIR/preprocesscli"
echo " ✓ preprocesscli"
if [ "$WITH_PREPROCESS_MODELS" = "1" ]; then
for d in annotators dwpose depth_anything_v2_small; do
if [ -d "$PREPROCESS_MODEL_DIR/$d" ]; then
mkdir -p "$DIST_DIR/preprocess/$d"
cp -r "$PREPROCESS_MODEL_DIR/$d/." "$DIST_DIR/preprocess/$d/"
echo " ✓ preprocess/$d"
fi
done
echo " 提示: 远程用 --depth-model/--pose-det/--pose-est/--hed-model/--mlsd-model 指向 \$DIST/preprocess/..."
fi
fi
# ── 动态后端模式:复制 sd.cpp / ggml 共享库与后端插件 ──
if [ "$SD_BACKEND_DL" = "1" ]; then
echo ""
echo ">>> 复制 sd.cpp / ggml 共享库到 $DIST_DIR/lib"
SD_LIBS=(
"$SD_BUILD_DIR/bin/libstable-diffusion.so"
"$SD_BUILD_DIR/bin/libggml.so"
"$SD_BUILD_DIR/bin/libggml.so.0"
"$SD_BUILD_DIR/bin/libggml.so.0.15.3"
"$SD_BUILD_DIR/bin/libggml-base.so"
"$SD_BUILD_DIR/bin/libggml-base.so.0"
"$SD_BUILD_DIR/bin/libggml-base.so.0.15.3"
)
for f in "${SD_LIBS[@]}"; do
if [ -f "$f" ]; then
cp -L "$f" "$DIST_DIR/lib/"
echo " ✓ $(basename "$f")"
else
echo " ✗ $(basename "$f") 未找到"
fi
done
echo ""
echo ">>> 复制 CPU 后端插件到 $DIST_DIR"
for f in "$SD_BUILD_DIR/bin"/libggml-cpu-*.so; do
if [ -f "$f" ]; then
cp -L "$f" "$DIST_DIR/"
echo " ✓ $(basename "$f")"
fi
done
if [ "$WITH_CUDA_BACKEND" = "1" ]; then
echo ""
echo ">>> 复制 CUDA 后端插件到 $DIST_DIR"
if [ -f "$SD_BUILD_DIR/bin/libggml-cuda.so" ]; then
cp -L "$SD_BUILD_DIR/bin/libggml-cuda.so" "$DIST_DIR/"
echo " ✓ libggml-cuda.so"
else
echo " ✗ libggml-cuda.so 未找到"
fi
fi
fi
# 复制 workflow 示例(如果有)
for f in "$PROJECT_DIR"/workflow*.json; do
[ -f "$f" ] && cp "$f" "$DIST_DIR/"
done
# ── 创建远程启动脚本 ──
cat > "$DIST_DIR/run.sh" << 'RUNEOF'
#!/bin/bash
# run.sh — ComfyCLI 远程启动脚本
# 远程只需要 NVIDIA 驱动,无需 pip install torch
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
export LD_LIBRARY_PATH="$SCRIPT_DIR/lib:$SCRIPT_DIR:$LD_LIBRARY_PATH"
export GGML_BACKEND_PATH="$SCRIPT_DIR/libggml-cuda.so"
[ -d "$SCRIPT_DIR/florence2" ] && export FLORENCE2_MODEL_DIR="$SCRIPT_DIR/florence2"
[ -f "$SCRIPT_DIR/birefnet/onnx/model.onnx" ] && export BIREFNET_MODEL="$SCRIPT_DIR/birefnet/onnx/model.onnx"
[ -d "$SCRIPT_DIR/ocr" ] && export OCR_MODEL_DIR="$SCRIPT_DIR/ocr"
cd "$SCRIPT_DIR"
exec ./comfycli-bin "$@"
RUNEOF
chmod +x "$DIST_DIR/run.sh"
# ── 启动前检查脚本 ──
cat > "$DIST_DIR/check_env.sh" << 'CHECKEOF'
#!/bin/bash
# check_env.sh — 远程环境检查
echo "=== 环境检查 ==="
echo "GLIBC: $(ldd --version 2>&1 | head -1)"
echo "NVIDIA 驱动: $(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null || echo '未检测到')"
echo "CUDA 可用: $(nvidia-smi 2>/dev/null && echo '是' || echo '否')"
echo ""
echo "=== 依赖 .so 检查 ==="
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
for lib in libsdcpp_adapter.so libstable-diffusion.so libggml.so.0 libggml-base.so.0; do
f=$(find "$SCRIPT_DIR" -maxdepth 2 -name "$lib*" 2>/dev/null | head -1)
if [ -n "$f" ]; then echo " ✓ $lib"; else echo " ✗ $lib (未找到)"; fi
done
echo " CUDA 后端: $(find "$SCRIPT_DIR" -maxdepth 1 -name 'libggml-cuda.so*' 2>/dev/null | head -1 | xargs -r basename 2>/dev/null || echo '未找到')"
echo " LD_LIBRARY_PATH=$SCRIPT_DIR/lib:$SCRIPT_DIR"
CHECKEOF
chmod +x "$DIST_DIR/check_env.sh"
# ── 打包 tarball ──
echo ""
echo ">>> 打包部署包"
TARBALL="$DIST_DIR/comfycli_deploy${TARBALL_SUFFIX}.tar.gz"
TMP_TARBALL=$(mktemp -t comfycli_deploy.XXXXXX.tar.gz)
trap 'rm -f "$TMP_TARBALL"' EXIT
cd "$DIST_DIR"
tar czf "$TMP_TARBALL" --exclude="*.tar.gz" .
mv "$TMP_TARBALL" "$TARBALL"
trap - EXIT
echo " -> $TARBALL ($(du -h "$TARBALL" | cut -f1))"
echo ""
echo " 部署包目录: $DIST_DIR/"
echo " 文件:"
du -sh "$DIST_DIR"/* 2>/dev/null | grep -v '.tar.gz' | sed 's/^/ /'
# ── 可选: SCP 到远程 ──
if [ $# -ge 2 ] && [ "$1" = "--scp" ]; then
REMOTE="$2"
echo ""
echo ">>> SCP 到 $REMOTE:/opt/comfycli/"
ssh "$REMOTE" "mkdir -p /opt/comfycli" 2>/dev/null || true
scp "$TARBALL" "$REMOTE:/opt/comfycli/" 2>&1
ssh "$REMOTE" "cd /opt/comfycli && tar xzf $(basename "$TARBALL") && rm -f $(basename "$TARBALL")" 2>&1
echo " OK"
echo ""
echo "=== 远程命令 ==="
echo " 环境检查: ssh $REMOTE \"bash /opt/comfycli/check_env.sh\""
echo " 运行: ssh $REMOTE \"bash /opt/comfycli/run.sh workflow.json --output-dir ./output\""
fi
echo ""
echo "============================================"
echo " 完成"
echo "============================================"