202611150107_Krea2Edit双参考图_omnixpu加速_B580_16G_冷启
2026-11-15 01:07:48 · sess-20261115-010625 · 工作流 ComfyUI
success冷启omnixpu加速B58016G
总耗时
288.10s
采样耗时
141s
采样步数
8 / 8
每步采样
17.63s/it
模型载入
13.76GB
单次峰值
7.93GB
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 | Krea2-Turbo-MusebyStableYogi-V15-Pro-Q4_0_C73.gguf | #20 UnetLoaderGGUFAdvanced Krea2 | 8120.7 | 25.262 | 322 | 未识别 |
| 文本编码器 text_encoder | qwen3vl_4b_fp8_scaled.safetensors | #4 CLIPLoader Krea2TEModel_ | 4999.5 | 0.101 | 49500 | policy 相邻 |
| VAE vae | qwen_image_vae.safetensors | #5 VAELoaderKJ WanVAE | 484.1 | 0.200 | 2420 | 按角色匹配 |
| VAE vae | (工作流执行日志未记录文件名) | 图中无对应加载节点(间接加载) WanVAE | 484.1 | 1.913 | 253 | 未识别 |
| LoRA lora | Krea2Edit-identity_edit_v1_2.safetensors Krea2/Krea2Edit-identity_edit_v1_2.safetensors | #8 Power Lora Loader (rgthree) 强度 1 | — | — | — | 工作流结构 |
共 5 行:底模 1 · 文本编码器 1 · VAE 2 · LoRA 1。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 3 个关闭状态(on=false)的 LoRA 未计入:Krea2Edit-Anything2RealCharacters-V3.safetensors、Krea2Edit-Anime2Real.safetensors、krea2_turbo_lora_rank_64_bf16.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 3 个关闭状态(on=false)的 LoRA 未计入:Krea2Edit-Anything2RealCharacters-V3.safetensors、Krea2Edit-Anime2Real.safetensors、krea2_turbo_lora_rank_64_bf16.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) | 20 |
|---|---|
| 实际执行节点数(API 图) | 20 |
| 节点状态分布(UI 图 mode) | active 20 · muted 0 · bypassed 0 |
| 被 bypass 的节点 | 无 |
| 被 mute 的节点 | 无 |
节点详情 20 个节点 · 8 个插件
| 节点名 | 插件名 | 插件版本 |
|---|---|---|
CLIPLoader 加载CLIP text_encoders | comfy-core | 0.3.73 |
EmptySD3LatentImage 空Latent图像(SD3) | comfy-core | 0.33.2 |
KSampler K采样器 | comfy-core | 0.3.48 |
LoadImage 加载图像 | comfy-core | 0.27.1 |
LoadImage 加载图像 | comfy-core | 0.27.1 |
SaveImage 保存图像 | comfy-core | 0.26.1 |
VAEDecodeTiled VAE解码(分块) | comfy-core | 0.12.3 |
VAEEncodeTiled VAE编码(分块) | comfy-core | 0.15.1 |
VAEEncodeTiled VAE编码(分块) | comfy-core | 0.15.1 |
LogCollection | ComfyUI-FullLog | — |
UnetLoaderGGUFAdvanced | ComfyUI-GGUF | 6ea2651e7df6… |
VAELoaderKJ | comfyui-kjnodes | 50a0837f9aea… |
VRAM_Debug | comfyui-kjnodes | 988a877bba8d… |
Krea2EditGroundedEncode | comfyui-krea2edit lbouaraba | 86f886dac230… |
Krea2EditGroundedEncode | comfyui-krea2edit lbouaraba | 86f886dac230… |
Krea2EditModelPatch Krea2 Edit (source patch) | comfyui-krea2edit lbouaraba | 86f886dac230… |
LayerUtility: ImageScaleByAspectRatio V2 图层工具:按宽高比缩放 V2 | comfyui_layerstyle | d94bef1ee5ed… |
LayerUtility: ImageScaleByAspectRatio V2 图层工具:按宽高比缩放 V2 | comfyui_layerstyle | d94bef1ee5ed… |
VRAMCleanup | comfyui_memory_cleanup | 1.1.3 |
Power Lora Loader (rgthree) | rgthree-comfy | 27b4f4cdcf3b… |
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+6) |
| 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 | 0.45.2 preview=Latent2RGB |
| 插件数 | 68 |
| 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 0.2.41 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 | 23 |
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 |
|---|---|---|---|
| 01:07:48.433 | 0.00 | 0.00 | 下发存储策略 · qwen_image_vae.safetensors |
| 01:07:50.977 | 2.54 | 2.54 | 下发存储策略 · qwen3vl_4b_fp8_scaled.safetensors |
| 01:07:52.949 | 4.52 | 1.98 | 请求加载 Krea2TEModel_ |
| 01:07:53.050 | 4.62 | 0.10 | qwen3vl_4b_fp8_scaled.safetensors 就绪(4999.47 MB) |
| 01:09:11.660 | 83.23 | 78.61 | 请求加载 WanVAE |
| 01:09:11.860 | 83.43 | 0.20 | qwen_image_vae.safetensors 就绪(484.06 MB) |
| 01:09:24.435 | 96.00 | 12.57 | 请求加载 Krea2 |
| 01:09:49.697 | 121.26 | 25.26 | Krea2 就绪(8120.70 MB) |
| 01:12:24.943 | 276.51 | 155.25 | 请求加载 WanVAE |
| 01:12:26.856 | 278.42 | 1.91 | qwen_image_vae.safetensors 就绪(484.06 MB) |
| 01:12:36.326 | 287.89 | 9.47 | 执行完成 |
原始日志内嵌,不依赖 user/comfyui.log
来源
user/comfyui.log · 54 行 · 日志自报耗时 288.10 s原始日志全文
[2026-11-15 01:07:48.433] Model storage policy: fast_disk=False paths=['F:\\ComfyUI-aki-v3\\ComfyUI\\models\\vae\\qwen_image_vae.safetensors'][2026-11-15 01:07:48.439] Using pytorch attention in VAE[2026-11-15 01:07:48.441] Using pytorch attention in VAE[2026-11-15 01:07:49.120] VAE load device: xpu:0, offload device: cpu, dtype: torch.float32[2026-11-15 01:07:49.854] # 😺dzNodes: LayerStyle -> ImageScaleByAspectRatio V2 Processed 1 image(s).[2026-11-15 01:07:50.068] # 😺dzNodes: LayerStyle -> ImageScaleByAspectRatio V2 Processed 1 image(s).[2026-11-15 01:07:50.254] Found quantization metadata version 1[2026-11-15 01:07:50.255] Using MixedPrecisionOps for text encoder[2026-11-15 01:07:50.977] Model storage policy: fast_disk=False paths=['F:\\ComfyUI-aki-v3\\ComfyUI\\models\\text_encoders\\qwen3vl_4b_fp8_scaled.safetensors'][2026-11-15 01:07:52.949] Requested to load Krea2TEModel_[2026-11-15 01:07:53.050] loaded completely; 4999.47 MB loaded, full load: True[2026-11-15 01:07:53.075] CLIP/text encoder model load device: cpu, offload device: cpu, current: cpu, dtype: torch.float16[2026-11-15 01:09:11.660] Requested to load WanVAE[2026-11-15 01:09:11.860] loaded completely; 6515.95 MB usable, 484.06 MB loaded, full load: True[2026-11-15 01:09:21.256] gguf qtypes: Q4_0 (256), F32 (174)[2026-11-15 01:09:21.286] model weight dtype torch.bfloat16, manual cast: None[2026-11-15 01:09:21.290] model_type FLUX[2026-11-15 01:09:21.292] Model storage policy: fast_disk=False paths=[][2026-11-15 01:09:21.626] [OmniXPU] LoRA memory budget cached: source=1743.5MiB/512 tensors model_added=1743.5MiB model_budget=1743.5MiB clip_added=0.0MiB clip_budget=0.0MiB[2026-11-15 01:09:21.748] [krea2edit] pre-encoding sources at target 1024x704px (before sampling, fit_mode=fit)[2026-11-15 01:09:21.749] [krea2edit] _fit_encode_image: mode=fit in=(1, 1024, 704, 3) target_latent=128x88[2026-11-15 01:09:23.263] [krea2edit] _fit_encode_image: mode=fit in=(1, 752, 1024, 3) target_latent=128x88[2026-11-15 01:09:24.435] [OmniXPU] LoRA model budget applied: cached=1743.5MiB/512 tensors requested=991.2->2734.7MiB minimum=495.6->2239.1MiB xpu_memory=available xpu_allocated=505.4MiB xpu_reserved=2450.0MiB xpu_free=8228.2MiB xpu_total=11946.2MiB[2026-11-15 01:09:24.435] Requested to load Krea2[2026-11-15 01:09:49.697] loaded completely; 8683.08 MB usable, 8120.70 MB loaded, full load: True[2026-11-15 01:09:50.663] [OmniXPU] LoRA model load complete: xpu_memory=available xpu_allocated=7929.3MiB xpu_reserved=8120.0MiB xpu_free=2205.0MiB xpu_total=11946.2MiB[2026-11-15 01:09:51.702] [krea2edit] pixel path ACTIVE (fit_mode=fit)[2026-11-15 01:09:52.330] [OmniXPU] norm first use: Krea2RMSNorm shape=torch.Size([843, 12, 2560])[2026-11-15 01:09:52.628] [OmniXPU] attention fallback: batch=843 (seq=12)[2026-11-15 01:09:55.639] [OmniXPU] attention fallback: batch=843 (seq=12)[2026-11-15 01:09:55.794] [OmniXPU] attention CUTE #1: heads=20 seq=843 dtype=torch.bfloat16[2026-11-15 01:09:55.968] [OmniXPU] attention CUTE #2: heads=20 seq=843 dtype=torch.bfloat16[2026-11-15 01:09:56.060] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:09:56.439] [OmniXPU] attention CUTE #3: heads=48 seq=7883 dtype=torch.bfloat16[2026-11-15 01:10:15.875] [OmniXPU] attention fallback: batch=843 (seq=12)[2026-11-15 01:10:15.951] [OmniXPU] attention fallback: batch=843 (seq=12)[2026-11-15 01:10:16.068] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:10:32.548] [OmniXPU] attention fallback: batch=843 (seq=12)[2026-11-15 01:10:32.743] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:10:49.356] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:11:06.246] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:11:22.976] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:11:39.623] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:11:56.247] [krea2edit] STRIDE1-POS fit: ref grids [(64, 44), (32, 44)] centered in (64,44)[2026-11-15 01:12:12.582] 100%|████████████████████████████████████████████████████████████████████████████████████| 8/8 [02:21<00:00, 16.86s/it]100%|████████████████████████████████████████████████████████████████████████████████████| 8/8 [02:21<00:00, 17.63s/it][2026-11-15 01:12:24.114] VRAM清理完成 [卸载模型: True, 清空缓存: True][2026-11-15 01:12:24.943] Requested to load WanVAE[2026-11-15 01:12:26.856] loaded completely; 10103.00 MB usable, 484.06 MB loaded, full load: True[2026-11-15 01:12:28.553] VRAMdebug: free memory before: 11,996,583,424[2026-11-15 01:12:29.130] VRAMdebug: free memory after: 12,526,496,768[2026-11-15 01:12:29.131] VRAMdebug: freed memory: 529,913,344[2026-11-15 01:12:36.326] Prompt executed in 288.10 seconds