Capturing execution frames: simple.json [ComfyUI] [comfy-env] [VRAM] Before offload 'VPRModel_2' (335 MB) -> cpu: 2169 / 7703 MB [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] call_id=31 sending model_to_device [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [VRAM] -219MB (now 910MB) reserved=1728MB peak=4556MB [ComfyUI] [comfy-env] [VRAM] After offload 'VPRModel_2' (335 MB): 1337 / 7703 MB [ComfyUI] [comfy-env] [VRAM] Before offload 'DepthAnything3Net_1' (783 MB) -> cpu: 1337 / 7703 MB [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] call_id=32 sending model_to_device [ComfyUI] [comfy-env] [VRAM] After offload 'DepthAnything3Net_1' (783 MB): 537 / 7703 MB Frame 1 saved (frame_000.jpg, t=0.0s) Validating workflow... [ComfyUI] === /validate endpoint called === [ComfyUI] Received JSON with keys: ['prompt'] [ComfyUI] Prompt has 4 nodes [ComfyUI] Node 2: LoadImage [ComfyUI] Inputs: {"image": "bridge.jpg"} Queuing workflow for execution... [ComfyUI] Node 3: DownloadAndLoadDepthAnythingV3Model [ComfyUI] Inputs: {"model": "da3_large.safetensors", "precision": "auto", "attention": "auto"} [ComfyUI] Node 4: PreviewImage [ComfyUI] Inputs: {"images": ["14", 0]} [ComfyUI] Node 14: DepthAnything_V3 [ComfyUI] Inputs: {"normalization_mode": "V2-Style", "resize_method": "resize", "invert_depth": true, "keep_model_size": false, "da3_model": ["3", 0], "images": ["2", 0]} [ComfyUI] Generated prompt_id: dfaac907-7eac-4cc0-9ecd-4e290975f0f8 [ComfyUI] Calling execution.validate_prompt()... [ComfyUI] Validation result: valid=True [ComfyUI] === Validation PASSED === [ComfyUI] got prompt [ComfyUI] [comfy-env] Running DownloadAndLoadDepthAnythingV3Model: model: 'da3_large.safetensors', precision: 'auto', attention: 'auto' [ComfyUI] [VRAM] Before DownloadAndLoadDepthAnythingV3Model: 537 / 7703 MB [ComfyUI] [SubprocessWorker] call_method: nodes.load_model.DownloadAndLoadDepthAnythingV3Model.EXECUTE_NORMALIZED [ComfyUI] [SubprocessWorker] acquired lock, ensuring started... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: failed [ComfyUI] [ComfyUI-DepthAnythingV3] Socket unhealthy, restarting worker... [ComfyUI] [comfy-env] Invalidated 2 stale model patchers (will be cleaned up during next unload) [ComfyUI] [ComfyUI-DepthAnythingV3] python: /root/.ce/.pixi/envs/depthanythingv3-nodes/bin/python [ComfyUI] [ComfyUI-DepthAnythingV3] sys_path sent to worker: ['/workspaces/DepthAnythingV3-0926/ComfyUI', '/workspaces/DepthAnythingV3-0926/ComfyUI/custom_nodes/ComfyUI-DepthAnythingV3', '/root/.ce/.pixi/envs/depthanythingv3-nodes/lib/python3.10/site-packages', '/workspaces/DepthAnythingV3-0926/ComfyUI/custom_nodes/ComfyUI-DepthAnythingV3'] [ComfyUI] [SubprocessWorker] is_pixi=True, python=/root/.ce/.pixi/envs/depthanythingv3-nodes/bin/python [ComfyUI] [SubprocessWorker] launching cmd=['/root/.pixi/bin/pixi', 'run', '--as-is', '--manifest-path', '/root/.ce/pixi.toml', '-e', 'depthanythingv3-nodes', 'python', '/tmp/comfyui_pvenv_xvidy7rt/persistent_worker.py', 'abstract://comfy_worker_c2d1059c5622']... [ComfyUI] [SubprocessWorker] socket_addr=abstract://comfy_worker_c2d1059c5622 (abstract, no filesystem path) [ComfyUI] [worker:ComfyUI-DepthAnythingV3] ERROR: Failed to import comfy_kitchen, Error: No module named 'comfy_kitchen', fp8 and fp4 support will not be available. [ComfyUI] [SubprocessWorker] worker started/confirmed [ComfyUI] [SubprocessWorker] serializing kwargs to shm... [ComfyUI] [SubprocessWorker] created 0 shm blocks [ComfyUI] [SubprocessWorker] sending request via socket... [ComfyUI] [SubprocessWorker] call_id=33 sending call_method [ComfyUI] [SubprocessWorker] got response: ok [ComfyUI] [comfy-env] [VRAM] Before offload 'VPRModel_2' (335 MB) -> cpu: 269 / 7703 MB [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] call_id=34 sending model_to_device [ComfyUI] Model 'VPRModel_2' gone from worker (restarted?), treating as offloaded [ComfyUI] [comfy-env] [VRAM] After offload 'VPRModel_2' (335 MB): 269 / 7703 MB [ComfyUI] [comfy-env] [VRAM] Before offload 'DepthAnything3Net_1' (783 MB) -> cpu: 269 / 7703 MB [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] call_id=35 sending model_to_device [ComfyUI] Model 'DepthAnything3Net_1' gone from worker (restarted?), treating as offloaded [ComfyUI] [comfy-env] [VRAM] After offload 'DepthAnything3Net_1' (783 MB): 269 / 7703 MB [ComfyUI] [comfy-env] DownloadAndLoadDepthAnythingV3Model done (1.75s): result: NodeOutput [ComfyUI] [VRAM] After DownloadAndLoadDepthAnythingV3Model: 269 / 7703 MB [ComfyUI] [comfy-env] Running DepthAnything_V3: normalization_mode: 'V2-Style', resize_method: 'resize', invert_depth: True, keep_model_size: False, da3_model: dict, images: torch.float32 [1,683,1024,3] cpu [ComfyUI] [VRAM] Before DepthAnything_V3: 269 / 7703 MB [ComfyUI] [SubprocessWorker] call_method: nodes.nodes_inference.DepthAnything_V3.EXECUTE_NORMALIZED [ComfyUI] [SubprocessWorker] acquired lock, ensuring started... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] worker started/confirmed [ComfyUI] [comfy-env] PRE-SERIALIZE 'images' shape: torch.Size([1, 683, 1024, 3]) [ComfyUI] [SubprocessWorker] serializing kwargs to shm... [ComfyUI] [SubprocessWorker] created 0 shm blocks [ComfyUI] [SubprocessWorker] sending request via socket... [ComfyUI] [SubprocessWorker] call_id=36 sending call_method [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [comfy-env] _from_shm got dict with keys: ['__type__', 'strategy', 'parent_pid', 'fd', 'storage_size'] [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [comfy-env] DESERIALIZE TensorRef: tensor_size=[1, 683, 1024, 3] [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [comfy-env] DESERIALIZED tensor shape: torch.Size([1, 683, 1024, 3]) [ComfyUI] [worker:ComfyUI-DepthAnythingV3] WARNING: No comfy kitchen, using old apply_rope functions. Frame 2 saved (frame_001.jpg, t=2.4s) [capture-periodic] freeze=1ms shot=128ms unfreeze=2ms saved=True [ComfyUI] [SubprocessWorker] callback 'request_vram_budget' from call_id=36 [ComfyUI] [comfy-env] VRAM request: 0.82GB, free before eviction: 7.80GB [ComfyUI] [comfy-env] VRAM after eviction: 7.80GB [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [VRAM] +234MB (now 234MB) reserved=256MB peak=234MB [ComfyUI] [worker:ComfyUI-DepthAnythingV3] WARNING: WARNING: This model does not support sky segmentation. V2-Style normalization will work but without sky masking. Use Mono/Metric/Nested models for best V2-Style results. Frame 3 saved (frame_002.jpg, t=4.6s) [capture-periodic] freeze=1ms shot=139ms unfreeze=2ms saved=True [ComfyUI] [worker:ComfyUI-DepthAnythingV3] [VRAM] +577MB (now 812MB) reserved=2560MB peak=1505MB [ComfyUI] [SubprocessWorker] callback 'report_progress' from call_id=36 [ComfyUI] [SubprocessWorker] got response: ok [ComfyUI] [comfy-env] Created 1 model patchers: ['DepthAnything3Net_1'] [ComfyUI] [comfy-env] DepthAnything_V3 done (3.08s): result: NodeOutput [ComfyUI] [VRAM] After DepthAnything_V3: 2937 / 7703 MB [ComfyUI] [VRAM] Loaded: SubprocessModel (783 MB) [ComfyUI] Prompt executed in 4.89 seconds [ComfyUI] [comfy-env] [VRAM] Before offload 'DepthAnything3Net_1' (783 MB) -> cpu: 2937 / 7703 MB [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ping (timeout=5.0s)... [ComfyUI] [ComfyUI-DepthAnythingV3] Health check: ok [ComfyUI] [SubprocessWorker] call_id=37 sending model_to_device Node executed (1 total), capturing... Frame 4 saved (frame_003.jpg, t=5.3s) [capture-node] freeze=2ms shot=150ms unfreeze=0ms saved=True Execution complete (t=5.3s) Captured 4 unique frames over 5.5s [debug] iframes: 0 [timing] trigger_3d_previews: 0.0s [ComfyUI] [comfy-env] [VRAM] After offload 'DepthAnything3Net_1' (783 MB): 441 / 7703 MB Saved high-quality screenshot: simple_executed.png [timing] final_screenshot: 0.3s Captured 5 video frames