1786480171180|[INFO] setup plugin alembic.autogenerate.schemas 1786480171182|[INFO] setup plugin alembic.autogenerate.tables 1786480171183|[INFO] setup plugin alembic.autogenerate.types 1786480171184|[INFO] setup plugin alembic.autogenerate.constraints 1786480171185|[INFO] setup plugin alembic.autogenerate.defaults 1786480171187|[INFO] setup plugin alembic.autogenerate.comments 1786480171188|[INFO] setup plugin alembic.autogenerate.checkconstraint_byname 1786480171217|[INFO] Adding extra search path diffusion_models /home/inhohur/ai-hot/weights/minimax-h3/diffusion_models 1786480171218|[INFO] Adding extra search path text_encoders /home/inhohur/ai-hot/weights/minimax-h3/text_encoders 1786480171219|[INFO] Adding extra search path vae /home/inhohur/ai-hot/weights/minimax-h3/vae 1786480171221|[INFO] Adding extra search path loras /home/inhohur/ai-hot/weights/minimax-h3/loras 1786480171222|[INFO] Adding extra search path diffusion_models /home/inhohur/ai/weights/minimax-h3/diffusion_models 1786480171223|[INFO] Adding extra search path text_encoders /home/inhohur/ai/weights/minimax-h3/text_encoders 1786480171225|[INFO] Adding extra search path vae /home/inhohur/ai/weights/minimax-h3/vae 1786480171226|[INFO] Adding extra search path loras /home/inhohur/ai/weights/minimax-h3/loras 1786480171227|[INFO] Setting output directory to: /home/inhohur/ai/experiments/h3-r2v-speed/dockaiq-out 1786480171229|[INFO] Setting user directory to: /home/inhohur/ai/runtimes/comfyui-h3/user 1786480172714|[INFO] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_embedding', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']} 1786480172715|[INFO] Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'int8_linear', 'quantize_and_rotate_rowwise', 'quantize_int8_rowwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_']} 1786480172716|[INFO] Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']} 1786480172717|[INFO] Found comfy_kitchen backend hip: {'available': False, 'disabled': False, 'unavailable_reason': 'PyTorch ROCm/HIP runtime not available', 'capabilities': []} 1786480172718|[INFO] Checkpoint files will always be loaded safely. 1786480172979|[INFO] Total VRAM 124545 MB, total RAM 124545 MB 1786480172981|[INFO] pytorch version: 2.13.0+cu130 1786480172981|[INFO] Set vram state to: NORMAL_VRAM 1786480172982|[INFO] Device: cuda:0 NVIDIA GB10 : cudaMallocAsync 1786480172983|[INFO] Using async weight offloading with 2 streams 1786480172984|[INFO] Enabled pinned memory 112090.0 1786480173003|[INFO] Using pytorch attention 1786480174680|aimdo: /project/src-posix/cuda-funchooks.c:52:DEBUG:aimdo_setup_hooks: hooks successfully installed 1786480174681|aimdo: /project/src/control.c:254:INFO:comfy-aimdo integrated Linux GPU RAM headroom: 7784 MB 1786480174683|aimdo: /project/src/control.c:276:INFO:comfy-aimdo inited for GPU: NVIDIA GB10 (VRAM: 124545 MB) 1786480174684|[INFO] DynamicVRAM support detected and enabled 1786480174685|[INFO] Python version: 3.12.3 (main, Mar 3 2026, 12:15:18) [GCC 13.3.0] 1786480174686|[INFO] ComfyUI version: 0.30.2 1786480174687|[INFO] comfy-aimdo version: 0.4.11 1786480174689|[INFO] comfy-kitchen version: 0.2.26 1786480174690|[INFO] Setting temp directory to: /home/inhohur/ai/runtimes/comfyui-h3/temp/temp 1786480174691|[INFO] comfyui-frontend-package version: 1.47.12 1786480174692|[INFO] comfyui-workflow-templates version: 0.11.31 1786480174694|[INFO] comfyui-embedded-docs version: 0.5.9 1786480174695|[INFO] comfy-kitchen version: 0.2.26 1786480174697|[INFO] comfy-aimdo version: 0.4.11 1786480174698|[INFO] [Prompt Server] web root: /home/inhohur/ai/runtimes/comfyui-h3/.venv/lib/python3.12/site-packages/comfyui_frontend_package/static 1786480174700|[INFO] Asset seeder disabled 1786480174889|[INFO] No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate' 1786480176076|[INFO] ✓ Sol-Attn nodes loaded: ['SolAttnModelPatcher', 'SolAttnMiniMaxH3Patcher', 'SolAttnInfo', 'SolAttnConfig'] 1786480176077|[INFO] ✓ GPU: NVIDIA GB10 (SM121) — Sol-Attn via Triton JIT path (triton_ref) 1786480176078|[INFO] [H3-PROF] node timing installed (min=0.20s · H3_PROF=0 to disable) 1786480176079|[INFO] [H3-VAEBATCH] per-prompt stock reset installed 1786480176080|[INFO] 1786480176081|Import times for custom nodes: 1786480176082|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/websocket_image_save.py 1786480176083|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-H3-Prof 1786480176084|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-H3-VAEBatch 1786480176085|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-MiniMax-H3-Turbo 1786480176086|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-H3-FBC 1786480176087|[INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-KJNodes 1786480176088|[INFO] 0.6 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI_sol-attn_Blackwell 1786480176089|[INFO] 1786480176090|[INFO] Context impl SQLiteImpl. 1786480176091|[INFO] Will assume non-transactional DDL. 1786480176110|[INFO] Using RAM pressure cache. 1786480176111|[INFO] Starting server 1786480176112| 1786480176113|[INFO] To see the GUI go to: http://127.0.0.1:8194 1786480178797|[INFO] got prompt 1786480178853|[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float32 1786480178978|[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16 1786480178999|[INFO] Found quantization metadata version 1 1786480179000|[INFO] Using MixedPrecisionOps for text encoder 1786480179838|[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16 1786480179840|[INFO] [H3-PROF] node 2 CLIPLoader: 0.86s 1786480179860|[INFO] Requested to load MiniMaxH3VideoVAE 1786480179946|[INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. 1786480181561|[INFO] Requested to load MiniMaxH3TEModel_ 1786480181690|[INFO] Model MiniMaxH3TEModel_ prepared for dynamic VRAM loading. 14956MB Staged. 0 patches attached. Force pre-loaded 410 weights: 4572 KB. 1786480205291|[INFO] [H3-PROF] node 6 MiniMaxH3ReferenceToVideo: 25.45s 1786480205334|[INFO] Found quantization metadata version 1 1786480205336|[INFO] Detected mixed precision quantization 1786480205337|[INFO] Using mixed precision operations 1786480205338|[INFO] Native ops: int8_tensorwise, mxfp8, float8_e5m2, convrot_w4a4, nvfp4, float8_e4m3fn 1786480205354|[INFO] model weight dtype torch.bfloat16, manual cast: torch.bfloat16 1786480205357|[INFO] model_type FLOW 1786480205520|[MiniMaxH3TurboLoRA] pruned base [bypass]: lora=minimax_h3_fl2v_lightx2v_turbo_8step_v1.0_comfy.safetensors strength=1.0 | 416 backbone modules, 0 adapters bound, 0 bypass injections set | model=MiniMaxH3 weight_dtype=torch.bfloat16 weight_dev=cpu 1786480205522|[INFO] Applying MiniMax H3 Memory Efficient Sage Attention Patch to all transformer blocks 1786480205523|[INFO] [H3-FBC] installed on 50 blocks | threshold=0.23 start_step=2 1786480205584|[INFO] Requested to load MiniMaxH3 1786480205664|[INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. 1786480208217|[H3TURBO sampler] sigmas=[1.0, 0.9882, 0.973, 0.9524, 0.9231, 0.878, 0.8, 0.6316, 0.0] x.shape=(1, 1, 3593664) dtype=torch.float32 v_numel=3580416 a_numel=13248 shapes=[torch.Size([1, 24, 37, 84, 48]), torch.Size([1, 32, 2, 207])] 1786480208271| 0%| | 0/8 [00:00 lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=1000.00 video_rms=0.9994 audio_rms=0.9990 dtype=torch.bfloat16 1786480284889|[H3TURBO step 0] sv=1.0000->0.9882 denoised_rms=2.0316 video: x_rms=0.9880 v_rms=2.2568 audio: x_rms=0.9531 v_rms=4.1547 slope=4.0000 1786480284893| 12%|█▎ | 1/8 [01:16<08:56, 76.67s/it][H3TURBO fwd pruned/bypass] call#2 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=988.24 video_rms=0.9880 audio_rms=0.9531 dtype=torch.bfloat16 1786480356516|[H3TURBO step 1] sv=0.9882->0.9730 denoised_rms=1.1819 video: x_rms=0.9732 v_rms=1.5377 audio: x_rms=0.8996 v_rms=4.0307 slope=3.7319 1786480356519| 25%|██▌ | 2/8 [02:28<07:22, 73.70s/it][H3TURBO fwd pruned/bypass] call#3 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=972.97 video_rms=0.9732 audio_rms=0.8996 dtype=torch.bfloat16 1786480358737|[INFO] [H3-FBC] step 2: row_ratio=0.2590 (p50=0.2602 p90=0.2986 global=0.186309) -> compute 1786480428854|[H3TURBO step 2] sv=0.9730->0.9524 denoised_rms=1.1661 video: x_rms=0.9537 v_rms=1.5258 audio: x_rms=0.8364 v_rms=3.7293 slope=3.4225 1786480428857| 38%|███▊ | 3/8 [03:40<06:05, 73.08s/it][H3TURBO fwd pruned/bypass] call#4 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=952.38 video_rms=0.9537 audio_rms=0.8364 dtype=torch.bfloat16 1786480430390|[INFO] [H3-FBC] step 3: row_ratio=0.1715 (p50=0.1722 p90=0.1911 global=0.139066) -> skip 1786480430437|[H3TURBO step 3] sv=0.9524->0.9231 denoised_rms=1.1540 video: x_rms=0.9272 v_rms=1.5185 audio: x_rms=0.7603 v_rms=3.3131 slope=3.0625 1786480430440| 50%|█████ | 4/8 [03:42<02:59, 44.85s/it][H3TURBO fwd pruned/bypass] call#5 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=923.08 video_rms=0.9272 audio_rms=0.7603 dtype=torch.bfloat16 1786480431949|[INFO] [H3-FBC] step 4: row_ratio=0.1461 (p50=0.1460 p90=0.1580 global=0.123578) -> skip 1786480431998|[H3TURBO step 4] sv=0.9231->0.8780 denoised_rms=1.1406 video: x_rms=0.8891 v_rms=1.5095 audio: x_rms=0.6679 v_rms=2.8493 slope=2.6406 1786480432003| 62%|██████▎ | 5/8 [03:43<01:27, 29.24s/it][H3TURBO fwd pruned/bypass] call#6 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=878.05 video_rms=0.8891 audio_rms=0.6679 dtype=torch.bfloat16 1786480503583|[H3TURBO step 5] sv=0.8780->0.8000 denoised_rms=1.1450 video: x_rms=0.8329 v_rms=1.5130 audio: x_rms=0.5612 v_rms=2.1542 slope=2.1441 1786480505098| 75%|███████▌ | 6/8 [04:55<01:27, 43.64s/it][INFO] [H3-FBC] step 6: row_ratio=0.1910 (p50=0.1917 p90=0.2045 global=0.161467) -> skip 1786480505144|[H3TURBO step 6] sv=0.8000->0.6316 denoised_rms=1.1353 video: x_rms=0.7604 v_rms=1.5051 audio: x_rms=0.4485 v_rms=1.5844 slope=1.5625 1786480576716| 88%|████████▊ | 7/8 [04:56<00:29, 29.88s/it][H3TURBO step 7] sv=0.6316->0.0000 denoised_rms=1.1425 video: x_rms=1.1440 v_rms=1.5006 audio: x_rms=0.4646 v_rms=0.8243 slope=0.9025 1786480576717| 100%|██████████| 8/8 [06:08<00:00, 43.16s/it] 100%|██████████| 8/8 [06:08<00:00, 46.06s/it] 1786480576719|[INFO] [H3-PROF] node 11 SamplerCustomAdvanced: 371.19s 1786480576895|[INFO] Requested to load MiniMaxH3AudioVAE 1786480577469|[INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. 1786480578340|[INFO] [H3-PROF] node 13 VAEDecodeAudio: 1.45s 1786480578344|[INFO] [H3-VAEBATCH] tiled_decode -> batched 1786480578392|[INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. 1786480670914|[INFO] [H3-PROF] node 12 VAEDecode: 92.57s 1786480682335|[INFO] [H3-PROF] node 15 SaveVideo: 11.31s 1786480682337|[INFO] Prompt executed in 503.52 seconds