202611052314_ZIT文生图+一阶双采二阶单采
2026-11-05 23:14:55 · sess-20261105-231451 · 工作流 未命名工作流
success冷启omnixpu加速B58016G
总耗时
451.20s
采样耗时
259s
采样步数
15 / 15
每步采样
17.32s/it
模型载入
18.08GB
单次峰值
5.73GB
生成物

模型加载模型文件 / 加载节点来自工作流结构(API 图);载入 MB / 耗时 / MB/s / 路径来源来自执行日志
| 角色 | 模型文件 | 加载节点 | 载入 MB | 耗时 s | MB/s | 路径来源 |
|---|---|---|---|---|---|---|
| 底模 base | moodyWildMix_v20Distilled10STEPS.safetensors ext/moodyWildMix_v20Distilled10STEPS.safetensors | #97 DiffusionModelLoaderKJ Lumina2 | 5869.8 | 13.056 | 450 | policy 相邻 |
| 底模 base | z_image_turbo-Q4_K_M.gguf ZIT/z_image_turbo-Q4_K_M.gguf | #102 UnetLoaderGGUFAdvanced Lumina2 | 9668.2 2 次加载 | 17.296 | 559 | 未识别 |
| 文本编码器 text_encoder | Qwen3-4B-Q4_K_M.gguf | #100 CLIPLoaderGGUF ZImageTEModel_ | 2813.5 | 0.063 | 44659 | 未识别 |
| VAE vae | ae.safetensors | #82 VAELoader AutoencodingEngine | 159.9 | 0.965 | 166 | 未识别 |
| LoRA lora | ZIT_Kook_Zimage_瑶光.safetensors | #92 #93 Power Lora Loader (rgthree) 强度 0.5 | — | — | — | 工作流结构 |
| LoRA lora | ZIT_胸部调整_adazbsr_v2_0075.safetensors | #92 #93 #94 Power Lora Loader (rgthree) 强度 1 | — | — | — | 工作流结构 |
| LoRA lora | ZITZIB_DarkBeastZ8_SDA@Fok_extracted_lora_r256.safetensors | #92 Power Lora Loader (rgthree) 强度 0.5 | — | — | — | 工作流结构 |
| LoRA lora | ZITZIB_美术增强_Z-Image-Aesthetic-Base_v1.safetensors | #92 #94 Power Lora Loader (rgthree) 强度 #92 0.8 / #94 0.6 | — | — | — | 工作流结构 |
| LoRA lora | ZITZIB_增加细节_REDZ15_DetailDaemonZ_lora_v1.1.safetensors | #92 #94 Power Lora Loader (rgthree) 强度 0.5 | — | — | — | 工作流结构 |
| LoRA lora | ZITZIB_tongliya佟丽娅.safetensors | #92 #93 Power Lora Loader (rgthree) 强度 1 | — | — | — | 工作流结构 |
共 10 行:底模 2 · 文本编码器 1 · VAE 1 · LoRA 6。模型文件 / 加载节点来自工作流结构(API 图)。载入 MB / 耗时 / MB/s 来自工作流执行日志(`Requested to load` → `loaded completely`);同一文件一次执行里加载多次时合并为一行、数值是累计值,相加仍等于 KPI 的「模型载入」。LoRA 是给底模打补丁、不走模型加载,执行日志里没有它的载入数据,所以这三列留空——它的文件名只能来自结构。
工作流里另有 2 个关闭状态(on=false)的 LoRA 未计入:ZIT_honghanfu_lora_v1.safetensors、xyy1.0_10.safetensors。
LoRA 文件名按执行日志的模糊匹配结果改写过:Kook_Zimage_瑶光.safetensors → ZIT_Kook_Zimage_瑶光.safetensors。
MB/s 用于区分「内存回迁」(GB/s 级)与「磁盘重读」(数百 MB/s)。text_encoder 在某些栈上 device_to=cpu,它的加载耗时接近 0,不代表省了显存搬运。
工作流里另有 2 个关闭状态(on=false)的 LoRA 未计入:ZIT_honghanfu_lora_v1.safetensors、xyy1.0_10.safetensors。
LoRA 文件名按执行日志的模糊匹配结果改写过:Kook_Zimage_瑶光.safetensors → ZIT_Kook_Zimage_瑶光.safetensors。
MB/s 用于区分「内存回迁」(GB/s 级)与「磁盘重读」(数百 MB/s)。text_encoder 在某些栈上 device_to=cpu,它的加载耗时接近 0,不代表省了显存搬运。
工作流结构UI 图含 bypass 节点,API 图是真正执行的子集工作流导出
| 画布节点数(UI 图) | 27 |
|---|---|
| 实际执行节点数(API 图) | 21 |
| 节点状态分布(UI 图 mode) | active 27 · muted 0 · bypassed 0 |
| 被 bypass 的节点 | 无 |
| 被 mute 的节点 | 无 |
节点详情 21 个节点 · 10 个插件
| 节点名 | 插件名 | 插件版本 |
|---|---|---|
CLIPTextEncode CLIP Text Encode (Negative Prompt) | comfy-core | 0.3.75 |
CLIPTextEncode CLIP文本编码 | comfy-core | 0.3.75 |
KSampler K采样器 | comfy-core | 0.9.1 |
KSamplerAdvanced K采样器(高级) | comfy-core | 0.3.75 |
KSamplerAdvanced K采样器(高级) | comfy-core | 0.3.75 |
LatentUpscaleBy 缩放Latent(比例) | comfy-core | 0.3.67 |
VAEDecodeTiled VAE解码(分块) | comfy-core | 0.18.1 |
VAELoader 加载VAE | comfy-core | 0.3.75 |
easy seed 随机种子 | comfyui-easy-use | 71c7865d2d3c… |
LogCollection | ComfyUI-FullLog | — |
CLIPLoaderGGUF | ComfyUI-GGUF | 6ea2651e7df6… |
UnetLoaderGGUFAdvanced | ComfyUI-GGUF-XPU analytics-zoo | 39671fe73117… |
Image Saver Metadata | comfyui-image-saver | a7c3d3766458… |
Image Saver Simple | comfyui-image-saver | a7c3d3766458… |
DiffusionModelLoaderKJ | comfyui-kjnodes | 50a0837f9aea… |
VRAM_Debug | comfyui-kjnodes | 9b565354e32a… |
CR Aspect Ratio | ComfyUI_Comfyroll_CustomNodes | d78b780ae43f… |
ImageSmartSharpen+ | comfyui_essentials | 9d9f4bedfc9f… |
Power Lora Loader (rgthree) | rgthree-comfy | 8ff50e452188… |
Power Lora Loader (rgthree) | rgthree-comfy | 8ff50e452188… |
Power Lora Loader (rgthree) | rgthree-comfy | 8ff50e452188… |
运行环境
| 系统 | Windows-11-10.0.22000-SP0 · AMD64 · build 10.0.22000 |
|---|---|
| 物理内存 | 15.9 GB (16G) |
| GPU | Intel(R) Arc(TM) B580 Graphics (0xE20B, 11946.2 MB, driver 1.17.39894+4) |
| 平台 | Intel(R) oneAPI Unified Runtime over Level-Zero V2 |
| 计算单元 | 160 |
| Python | 3.13.11 |
| torch | 2.15.0+xpu (xpu available: True, devices: 1) |
| ComfyUI | 0.44.0 preview=Latent2RGB |
| 插件数 | 67 |
| 加速栈 | |
| 加速方案 | omnixpu加速 (工作流执行日志中 omni_xpu_kernel 0.2.0b3+torch215.bmg 生效,且出现 [OmniXPU] 插件标记) |
| OmniXPU 内核 | omni_xpu_kernel 0.2.0b3+torch215.bmg 已生效:sdp, norm, rotary, linear_fp8, int8, layout |
| OmniXPU 插件 | ComfyUI-OmniXPU 16 个 adapter applied、4 个 skipped |
| aimdo | comfy-aimdo 0.6.0 未启用 (disabled by env) |
| kitchen XPU | comfy-kitchen 0.2.40 xpu 后端可用 (adaln, apply_rope, apply_rope1, apply_rope1_… 共 50 项) |
加速方案 = 内核生效行 + 插件
[OmniXPU] 标记,两处都在工作流执行日志里才判「omnixpu加速」;aimdo / kitchen 是可选栈,不影响方案判定,也不进文件名。运行时环境变量 18 项
| 变量名 | 值 |
|---|---|
OCL_ICD_FILENAMES | ;F:\Intel-oneAPI\compiler\latest\bin\intelocl64.dll |
OMNIXPU_DEBUG | 0 |
OMNIXPU_ENABLE | 1 |
OMNIXPU_QWEN_IMAGE21_QKV | 1 |
ONEAPI_CLANG_VERSION | 22 |
ONEAPI_DEVICE_SELECTOR | level_zero:0 |
ONEAPI_PATH | F:\Intel-oneAPI |
ONEAPI_ROOT | F:\Intel-oneAPI |
SYCL_CACHE_MAX_SIZE | 8192 |
SYCL_CACHE_THRESHOLD | 7 |
SYCL_DEVICE_FILTER | level_zero |
SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS | 0 |
SYCL_PI_LEVEL_ZERO_TRACK_INDIRECT_ACCESS_MEMORY | 1 |
SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS | 1 |
UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS | 1 |
UR_L0_USE_IMMEDIATE_COMMANDLISTS | 1 |
ZE_DEVICE_SLEEP | 0 |
ZE_ENABLE_PCI_ID_DEVICE_ORDER | 1 |
时间轴
| 时刻 | 相对起点 s | 增量 s | 事件 |
|---|---|---|---|
| 23:14:55.494 | 0.00 | 0.00 | 下发存储策略 · ae.safetensors |
| 23:15:08.276 | 12.78 | 12.78 | 请求加载 ZImageTEModel_ |
| 23:15:08.339 | 12.85 | 0.07 | ZImageTEModel_ 就绪(2813.50 MB) |
| 23:15:08.367 | 12.87 | 0.02 | 请求加载 ZImageTEModel_ |
| 23:15:51.673 | 56.18 | 43.31 | 请求加载 Lumina2 |
| 23:16:07.027 | 71.53 | 15.35 | ae.safetensors 就绪(4834.06 MB) |
| 23:16:24.999 | 89.50 | 17.97 | 下发存储策略 · moodyWildMix_v20Distilled10STEPS.safetensors |
| 23:16:58.402 | 122.91 | 33.41 | 请求加载 Lumina2 |
| 23:17:11.458 | 135.96 | 13.05 | ae.safetensors 就绪(5869.77 MB) |
| 23:17:22.262 | 146.77 | 10.81 | 请求加载 Lumina2 |
| 23:17:24.204 | 148.71 | 1.94 | ae.safetensors 就绪(4834.06 MB) |
| 23:21:44.502 | 409.01 | 260.30 | 请求加载 AutoencodingEngine |
| 23:21:45.467 | 409.97 | 0.96 | ae.safetensors 就绪(159.87 MB) |
| 23:22:26.591 | 451.10 | 41.13 | 执行完成 |
原始日志内嵌,不依赖 user/comfyui.log
来源
user/comfyui.log · 67 行 · 日志自报耗时 451.20 s原始日志全文
[2026-11-05 23:14:55.494] Model storage policy: fast_disk=False paths=['F:\\ComfyUI-aki-v3\\ComfyUI\\models\\vae\\ae.safetensors'][2026-11-05 23:14:55.500] Using pytorch attention in VAE[2026-11-05 23:14:55.501] Using pytorch attention in VAE[2026-11-05 23:14:56.474] VAE load device: xpu:0, offload device: cpu, dtype: torch.bfloat16[2026-11-05 23:15:05.423] gguf qtypes: F32 (145), Q6_K (37), Q4_K (216)[2026-11-05 23:15:05.568] Dequantizing token_embd.weight to prevent runtime OOM.[2026-11-05 23:15:08.226] Model storage policy: fast_disk=False paths=[][2026-11-05 23:15:08.276] Requested to load ZImageTEModel_[2026-11-05 23:15:08.339] loaded completely; 2813.50 MB loaded, full load: True[2026-11-05 23:15:08.354] CLIP/text encoder model load device: cpu, offload device: cpu, current: cpu, dtype: torch.float16[2026-11-05 23:15:08.367] Requested to load ZImageTEModel_[2026-11-05 23:15:50.950] gguf qtypes: F32 (245), BF16 (28), Q4_K (120), Q5_K (30), Q6_K (30)[2026-11-05 23:15:50.990] model weight dtype torch.bfloat16, manual cast: None[2026-11-05 23:15:50.994] model_type FLOW[2026-11-05 23:15:50.996] Model storage policy: fast_disk=False paths=[][2026-11-05 23:15:51.110] [rgthree-comfy][Power Lora Loader] Fuzzy-matched Lora input "Kook_Zimage_瑶光.safetensors" to "ZIT_Kook_Zimage_瑶光.safetensors".[2026-11-05 23:15:51.330] [OmniXPU] LoRA memory budget cached: source=302.8MiB/476 tensors model_added=302.8MiB model_budget=302.8MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:15:51.469] [OmniXPU] LoRA memory budget cached: source=40.5MiB/480 tensors model_added=40.5MiB model_budget=343.4MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:15:51.609] [OmniXPU] LoRA memory budget cached: source=40.0MiB/630 tensors model_added=39.8MiB model_budget=383.2MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:15:51.673] [OmniXPU] LoRA model budget applied: cached=383.2MiB/1376 tensors requested=1806.3->2189.5MiB minimum=903.2->1286.3MiB xpu_memory=available xpu_allocated=0.0MiB xpu_reserved=2.0MiB xpu_free=10342.1MiB xpu_total=11946.2MiB[2026-11-05 23:15:51.673] Requested to load Lumina2[2026-11-05 23:16:07.027] loaded completely; 9635.86 MB usable, 4834.06 MB loaded, full load: True[2026-11-05 23:16:07.496] [OmniXPU] LoRA model load complete: xpu_memory=available xpu_allocated=4846.0MiB xpu_reserved=4942.0MiB xpu_free=5503.7MiB xpu_total=11946.2MiB[2026-11-05 23:16:08.730] [OmniXPU] norm first use: RMSNorm shape=torch.Size([1, 308, 2560])[2026-11-05 23:16:09.215] [OmniXPU] attention CUTE #1: heads=30 seq=320 dtype=torch.bfloat16[2026-11-05 23:16:09.439] [OmniXPU] attention CUTE #2: heads=30 seq=320 dtype=torch.bfloat16[2026-11-05 23:16:09.644] [OmniXPU] attention CUTE #3: heads=30 seq=4032 dtype=torch.bfloat16[2026-11-05 23:16:24.782] 100%|████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:16<00:00, 2.96s/it]100%|████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:16<00:00, 3.32s/it][2026-11-05 23:16:24.804] Setting ext\moodyWildMix_v20Distilled10STEPS.safetensors weight dtype to torch.float8_e4m3fn[2026-11-05 23:16:24.995] model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16[2026-11-05 23:16:24.998] model_type FLOW[2026-11-05 23:16:24.999] Model storage policy: fast_disk=False paths=['E:\\model\\moodyWildMix_v20Distilled10STEPS.safetensors'][2026-11-05 23:16:57.577] Setting ext\moodyWildMix_v20Distilled10STEPS.safetensors compute dtype to torch.float16[2026-11-05 23:16:57.911] [OmniXPU] LoRA memory budget cached: source=40.5MiB/480 tensors model_added=40.5MiB model_budget=40.5MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:16:58.121] [OmniXPU] LoRA memory budget cached: source=133.8MiB/630 tensors model_added=133.6MiB model_budget=174.1MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:16:58.299] [OmniXPU] LoRA memory budget cached: source=351.7MiB/450 tensors model_added=351.6MiB model_budget=525.7MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:16:58.399] [OmniXPU] LoRA model budget applied: cached=525.7MiB/1200 tensors requested=1806.3->2332.0MiB minimum=903.2->1428.9MiB xpu_memory=available xpu_allocated=4846.0MiB xpu_reserved=5826.0MiB xpu_free=4349.3MiB xpu_total=11946.2MiB[2026-11-05 23:16:58.402] Requested to load Lumina2[2026-11-05 23:17:00.458] Unloaded partially: 2835.43 MB freed, 1998.65 MB remains loaded, 92.29 MB buffer reserved, lowvram patches: 162[2026-11-05 23:17:11.458] loaded completely; 7567.05 MB usable, 5869.77 MB loaded, full load: True[2026-11-05 23:17:11.466] [OmniXPU] LoRA model load complete: xpu_memory=available xpu_allocated=7845.4MiB xpu_reserved=7944.0MiB xpu_free=1672.6MiB xpu_total=11946.2MiB[2026-11-05 23:17:20.973] 100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:09<00:00, 1.14it/s]100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:09<00:00, 1.05it/s][2026-11-05 23:17:21.469] [rgthree-comfy][Power Lora Loader] Fuzzy-matched Lora input "Kook_Zimage_瑶光.safetensors" to "ZIT_Kook_Zimage_瑶光.safetensors".[2026-11-05 23:17:21.689] [OmniXPU] LoRA memory budget cached: source=302.8MiB/476 tensors model_added=302.8MiB model_budget=302.8MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:21.751] [OmniXPU] LoRA memory budget cached: source=40.5MiB/480 tensors model_added=40.5MiB model_budget=343.4MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:21.921] [OmniXPU] LoRA memory budget cached: source=1083.8MiB/340 tensors model_added=1083.8MiB model_budget=1427.1MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:21.996] [OmniXPU] LoRA memory budget cached: source=351.7MiB/450 tensors model_added=351.6MiB model_budget=1778.7MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:22.073] [OmniXPU] LoRA memory budget cached: source=133.8MiB/630 tensors model_added=133.6MiB model_budget=1912.3MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:22.226] [OmniXPU] LoRA memory budget cached: source=40.0MiB/630 tensors model_added=39.8MiB model_budget=1952.1MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-05 23:17:22.261] [OmniXPU] LoRA model budget applied: cached=1952.1MiB/2436 tensors requested=7225.3->9177.4MiB minimum=3612.7->5564.8MiB xpu_memory=available xpu_allocated=1926.3MiB xpu_reserved=8756.0MiB xpu_free=844.5MiB xpu_total=11946.2MiB[2026-11-05 23:17:22.262] Requested to load Lumina2[2026-11-05 23:17:24.204] loaded completely; 5429.82 MB usable, 4834.06 MB loaded, full load: True[2026-11-05 23:17:24.511] [OmniXPU] LoRA model load complete: xpu_memory=available xpu_allocated=4846.0MiB xpu_reserved=4942.0MiB xpu_free=5010.5MiB xpu_total=11946.2MiB[2026-11-05 23:21:44.476] 100%|██████████████████████████████████████████████████████████████████████████████████| 15/15 [04:19<00:00, 17.01s/it]100%|██████████████████████████████████████████████████████████████████████████████████| 15/15 [04:19<00:00, 17.32s/it][2026-11-05 23:21:44.502] Requested to load AutoencodingEngine[2026-11-05 23:21:45.467] loaded completely; 5257.02 MB usable, 159.87 MB loaded, full load: True[2026-11-05 23:22:05.800] VRAMdebug: free memory before: 7,275,137,024[2026-11-05 23:22:14.406] VRAMdebug: free memory after: 12,526,496,768[2026-11-05 23:22:14.407] VRAMdebug: freed memory: 5,251,359,744[2026-11-05 23:22:15.215] The path `F:\ComfyUI-aki-v3\ComfyUI\output\2026-11-05` specified doesn't exist! Creating directory.[2026-11-05 23:22:26.591] Prompt executed in 451.20 seconds