202611030246_QwenImage2.1文生图_viggle
2026-11-03 02:46:35 · sess-20261103-024608 · 工作流 ComfyUI
success冷启omnixpu加速B58016G
总耗时
193.61s
采样耗时
7s
采样步数
4 / 4
每步采样
1.92s/it
模型载入
12.41GB
单次峰值
7.75GB
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 | qwen_image_2.1-viggle-4-steps-turbo_Q4_K_M.gguf ext/qwen_image_2.1-viggle-4-steps-turbo_Q4_K_M.gguf | #1 UnetLoaderGGUFAdvanced QwenImage21 | 4123.2 | 36.451 | 113 | 未识别 |
| 文本编码器 text_encoder | Qwen3VL-8B-Instruct-Q4_K_M.gguf ext/Qwen3VL-8B-Instruct-Q4_K_M.gguf | #2 CLIPLoaderGGUF QwenImage21TEModel_ | 7938.0 | 0.125 | 63504 | 未识别 |
| VAE vae | qwen_image_2.1_vae_bf16.safetensors | #3 VAELoaderKJ WanVAE | 644.2 | 2.726 | 236 | 按角色匹配 |
共 3 行(图中每一处模型引用各占一行,不合并):3 行在执行日志里有载入数据,0 行只有结构引用。
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。
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导出
| 画布节点数(UI 图) | 12 |
|---|---|
| 实际执行节点数(API 图) | 12 |
| 节点状态分布(UI 图 mode) | active 12 · muted 0 · bypassed 0 |
| 被 bypass 的节点 | 无 |
| 被 mute 的节点 | 无 |
节点详情 12 个节点 · 6 个插件
| 节点名 | 插件名 | 插件版本 |
|---|---|---|
EmptyLatentImage 空Latent图像 | comfy-core | 0.41.0 |
KSampler K采样器 | comfy-core | 0.37.0 |
ResolutionSelector 分辨率选择器 | comfy-core | 0.41.0 |
SaveImage 保存图像 | comfy-core | 0.3.48 |
TextEncodeQwenImage21 文本编码 Qwen Image 2.1 | comfy-core | 0.37.0 |
VAEDecode VAE解码 | comfy-core | 0.37.0 |
LogCollection | ComfyUI-FullLog | — |
CLIPLoaderGGUF | ComfyUI-GGUF | 6ea2651e7df6… |
UnetLoaderGGUFAdvanced | ComfyUI-GGUF-XPU analytics-zoo | 39671fe73117… |
VAELoaderKJ | comfyui-kjnodes | 50a0837f9aea… |
VRAM_Debug | comfyui-kjnodes | 988a877bba8d… |
VRAMCleanup | comfyui_memory_cleanup | 1.1.3 |
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 | 0.43.1 preview=Latent2RGB |
| 插件数 | 67 |
| 加速栈 | |
| 加速方案 | omnixpu加速 (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 15 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 17 项
| 变量名 | 值 |
|---|---|
OCL_ICD_FILENAMES | ;F:\Intel-oneAPI\compiler\latest\bin\intelocl64.dll |
OMNIXPU_DEBUG | 0 |
OMNIXPU_ENABLE | 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 |
|---|---|---|---|
| 02:46:35.818 | 0.13 | 0.13 | 下发存储策略 · qwen_image_2.1_vae_bf16.safetensors |
| 02:46:56.903 | 21.21 | 21.08 | 请求加载 QwenImage21TEModel_ |
| 02:46:57.028 | 21.34 | 0.13 | QwenImage21TEModel_ 就绪(7938.04 MB) |
| 02:46:57.076 | 21.38 | 0.04 | 请求加载 QwenImage21TEModel_ |
| 02:48:30.146 | 114.45 | 93.07 | 请求加载 QwenImage21 |
| 02:49:06.597 | 150.91 | 36.46 | QwenImage21 就绪(4123.23 MB) |
| 02:49:26.996 | 171.30 | 20.39 | 请求加载 WanVAE |
| 02:49:29.722 | 174.03 | 2.73 | qwen_image_2.1_vae_bf16.safetensors 就绪(644.19 MB) |
| 02:49:49.248 | 193.56 | 19.53 | 执行完成 |
原始日志内嵌,不依赖 user/comfyui.log
来源
user/comfyui.log · 41 行 · 日志自报耗时 193.61 s原始日志全文
[2026-11-03 02:46:35.692] gguf qtypes: BF16 (3), Q4_K (197), F32 (65), Q6_K (32)[2026-11-03 02:46:35.712] model weight dtype torch.bfloat16, manual cast: None[2026-11-03 02:46:35.727] model_type FLUX[2026-11-03 02:46:35.728] Model storage policy: fast_disk=False paths=[][2026-11-03 02:46:35.818] Model storage policy: fast_disk=False paths=['F:\\ComfyUI-aki-v3\\ComfyUI\\models\\vae\\qwen_image_2.1_vae_bf16.safetensors'][2026-11-03 02:46:35.828] Using pytorch attention in VAE[2026-11-03 02:46:35.831] Using pytorch attention in VAE[2026-11-03 02:46:37.835] VAE load device: xpu:0, offload device: cpu, dtype: torch.bfloat16[2026-11-03 02:46:46.646] gguf qtypes: Q6_K (37), F32 (145), Q4_K (217)[2026-11-03 02:46:46.778] Dequantizing token_embd.weight to prevent runtime OOM.[2026-11-03 02:46:56.283] Attenpting to find mmproj file for text encoder...[2026-11-03 02:46:56.284] Using mmproj 'mmproj-Qwen3VL-8B-Instruct-F16.gguf' for text encoder 'Qwen3VL-8B-Instruct-Q4_K_M.gguf'.[2026-11-03 02:46:56.318] gguf qtypes: F32 (234), F16 (118)[2026-11-03 02:46:56.336] Qwen3-VL DeepStack: remapped vision layers [8, 16, 24] -> list index [0, 1, 2].[2026-11-03 02:46:56.750] Model storage policy: fast_disk=False paths=[][2026-11-03 02:46:56.903] Requested to load QwenImage21TEModel_[2026-11-03 02:46:57.028] loaded completely; 7938.04 MB loaded, full load: True[2026-11-03 02:46:57.063] CLIP/text encoder model load device: cpu, offload device: cpu, current: cpu, dtype: torch.float16[2026-11-03 02:46:57.076] Requested to load QwenImage21TEModel_[2026-11-03 02:48:30.146] Requested to load QwenImage21[2026-11-03 02:49:06.597] loaded completely; 10103.00 MB usable, 4123.23 MB loaded, full load: True[2026-11-03 02:49:07.658] [OmniXPU] norm first use: rms_norm_fn shape=torch.Size([1, 226, 4096])[2026-11-03 02:49:09.149] [OmniXPU] attention fallback: mask=torch.Size([226, 226]) (seq=226)[2026-11-03 02:49:09.503] [OmniXPU] experimental capability-driven BMG D128 CUTE route enabled; set OMNI_ATTN_BACKEND=torch to disable it if this workflow shows an issue[2026-11-03 02:49:09.503] [OmniXPU] attention CUTE_BHLD_D128 #1: heads=32 q=4941 kv=5167 dtype=torch.bfloat16[2026-11-03 02:49:09.637] [OmniXPU] attention fallback: mask=torch.Size([226, 226]) (seq=226)[2026-11-03 02:49:09.643] [OmniXPU] attention CUTE_BHLD_D128 #2: heads=32 q=4941 kv=5167 dtype=torch.bfloat16[2026-11-03 02:49:09.648] [OmniXPU] attention fallback: mask=torch.Size([226, 226]) (seq=226)[2026-11-03 02:49:09.654] [OmniXPU] attention CUTE_BHLD_D128 #3: heads=32 q=4941 kv=5167 dtype=torch.bfloat16[2026-11-03 02:49:09.659] [OmniXPU] attention fallback: mask=torch.Size([226, 226]) (seq=226)[2026-11-03 02:49:09.669] [OmniXPU] attention fallback: mask=torch.Size([226, 226]) (seq=226)[2026-11-03 02:49:15.300] 100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:07<00:00, 1.64s/it]100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:07<00:00, 1.92s/it][2026-11-03 02:49:26.724] VRAM清理完成 [卸载模型: True, 清空缓存: True][2026-11-03 02:49:26.996] Requested to load WanVAE[2026-11-03 02:49:29.722] loaded completely; 8750.86 MB usable, 644.19 MB loaded, full load: True[2026-11-03 02:49:41.608] VRAMdebug: free memory before: 11,843,641,344[2026-11-03 02:49:42.275] VRAMdebug: free memory after: 12,526,496,768[2026-11-03 02:49:42.275] VRAMdebug: freed memory: 682,855,424[2026-11-03 02:49:49.248] Prompt executed in 193.61 seconds