[INFO] setup plugin alembic.autogenerate.schemas [INFO] setup plugin alembic.autogenerate.tables [INFO] setup plugin alembic.autogenerate.types [INFO] setup plugin alembic.autogenerate.constraints [INFO] setup plugin alembic.autogenerate.defaults [INFO] setup plugin alembic.autogenerate.comments [INFO] setup plugin alembic.autogenerate.checkconstraint_byname [INFO] Adding extra search path diffusion_models /home/inhohur/ai/weights/minimax-h3/diffusion_models [INFO] Adding extra search path text_encoders /home/inhohur/ai/weights/minimax-h3/text_encoders [INFO] Adding extra search path vae /home/inhohur/ai/weights/minimax-h3/vae [INFO] Adding extra search path loras /home/inhohur/ai/weights/minimax-h3/loras [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']} [INFO] Found comfy_kitchen backend hip: {'available': False, 'disabled': False, 'unavailable_reason': 'PyTorch ROCm/HIP runtime not available', 'capabilities': []} [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_']} [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']} [INFO] Checkpoint files will always be loaded safely. [INFO] Total VRAM 124545 MB, total RAM 124545 MB [INFO] pytorch version: 2.13.0+cu130 [INFO] Set vram state to: NORMAL_VRAM [INFO] Device: cuda:0 NVIDIA GB10 : cudaMallocAsync [INFO] Using async weight offloading with 2 streams [INFO] Enabled pinned memory 112090.0 [INFO] Using pytorch attention aimdo: /project/src-posix/cuda-funchooks.c:52:DEBUG:aimdo_setup_hooks: hooks successfully installed aimdo: /project/src/control.c:254:INFO:comfy-aimdo integrated Linux GPU RAM headroom: 7784 MB aimdo: /project/src/control.c:276:INFO:comfy-aimdo inited for GPU: NVIDIA GB10 (VRAM: 124545 MB) [INFO] DynamicVRAM support detected and enabled [INFO] Python version: 3.12.3 (main, Mar 3 2026, 12:15:18) [GCC 13.3.0] [INFO] ComfyUI version: 0.30.2 [INFO] comfy-aimdo version: 0.4.11 [INFO] comfy-kitchen version: 0.2.26 [WARNING] ****** User settings have been changed to be stored on the server instead of browser storage. ****** [WARNING] ****** For multi-user setups add the --multi-user CLI argument to enable multiple user profiles. ****** [INFO] comfyui-frontend-package version: 1.47.12 [INFO] comfyui-workflow-templates version: 0.11.31 [INFO] comfyui-embedded-docs version: 0.5.9 [INFO] comfy-kitchen version: 0.2.26 [INFO] comfy-aimdo version: 0.4.11 [INFO] [Prompt Server] web root: /home/inhohur/ai/runtimes/comfyui-h3/.venv/lib/python3.12/site-packages/comfyui_frontend_package/static [INFO] Asset seeder disabled [INFO] No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate' [INFO] Import times for custom nodes: [INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/websocket_image_save.py [INFO] 0.0 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-MiniMax-H3-Turbo [INFO] 0.2 seconds: /mnt/ai/runtimes/comfyui-h3/repo/custom_nodes/ComfyUI-KJNodes [INFO] [INFO] Context impl SQLiteImpl. [INFO] Will assume non-transactional DDL. [INFO] Context impl SQLiteImpl. [INFO] Will assume non-transactional DDL. [INFO] Running upgrade -> 0001_assets, Initial assets schema Revision ID: 0001_assets Revises: None Create Date: 2025-12-10 00:00:00 [INFO] Running upgrade 0001_assets -> 0002_merge_to_asset_references, Merge AssetInfo and AssetCacheState into unified asset_references table. [INFO] Running upgrade 0002_merge_to_asset_references -> 0003_add_metadata_job_id, Add system_metadata and job_id columns to asset_references. Change preview_id FK from assets.id to asset_references.id. [INFO] Running upgrade 0003_add_metadata_job_id -> 0004_drop_tag_type, Drop the vestigial tags.tag_type column. [INFO] Running upgrade 0004_drop_tag_type -> 0005_allow_case_sensitive_tags, Allow case-sensitive tag names. [INFO] Running upgrade 0005_allow_case_sensitive_tags -> 0006_add_loader_path, Add loader_path column to asset_references. [INFO] Database upgraded from None to 0006_add_loader_path [INFO] Using RAM pressure cache. [INFO] Starting server [INFO] To see the GUI go to: http://127.0.0.1:8190 [INFO] got prompt [INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float32 [INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16 [INFO] Found quantization metadata version 1 [INFO] Using MixedPrecisionOps for text encoder [INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16 [INFO] Requested to load MiniMaxH3VideoVAE [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Requested to load MiniMaxH3TEModel_ [INFO] Model MiniMaxH3TEModel_ prepared for dynamic VRAM loading. 14956MB Staged. 0 patches attached. Force pre-loaded 410 weights: 4572 KB. [INFO] Found quantization metadata version 1 [INFO] Detected mixed precision quantization [INFO] Using mixed precision operations [INFO] Native ops: float8_e4m3fn, nvfp4, convrot_w4a4, float8_e5m2, mxfp8, int8_tensorwise [INFO] model weight dtype torch.bfloat16, manual cast: torch.bfloat16 [INFO] model_type FLOW [INFO] Applying MiniMax H3 Memory Efficient Sage Attention Patch to all transformer blocks [INFO] Requested to load MiniMaxH3 [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. 0%| | 0/20 [00:00 lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=1000.00 video_rms=0.9999 audio_rms=1.0110 dtype=torch.bfloat16 [H3TURBO step 0] sv=1.0000->0.9730 denoised_rms=2.0425 video: x_rms=0.9748 v_rms=2.2590 audio: x_rms=0.9089 v_rms=4.7463 slope=4.0000 25%|██▌ | 1/4 [01:09<03:27, 69.29s/it][H3TURBO fwd pruned/bypass] call#2 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=972.97 video_rms=0.9748 audio_rms=0.9089 dtype=torch.bfloat16 [H3TURBO step 1] sv=0.9730->0.9231 denoised_rms=1.1877 video: x_rms=0.9301 v_rms=1.5264 audio: x_rms=0.7665 v_rms=3.7680 slope=3.4225 50%|█████ | 2/4 [02:18<02:18, 69.17s/it][H3TURBO fwd pruned/bypass] call#3 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=923.08 video_rms=0.9301 audio_rms=0.7665 dtype=torch.bfloat16 [H3TURBO step 2] sv=0.9231->0.8000 denoised_rms=1.1655 video: x_rms=0.8393 v_rms=1.5116 audio: x_rms=0.5629 v_rms=2.8620 slope=2.6406 75%|███████▌ | 3/4 [03:28<01:09, 69.63s/it][H3TURBO fwd pruned/bypass] call#4 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=800.00 video_rms=0.8393 audio_rms=0.5629 dtype=torch.bfloat16 [H3TURBO step 3] sv=0.8000->0.0000 denoised_rms=1.1615 video: x_rms=1.1621 v_rms=1.5038 audio: x_rms=0.4520 v_rms=1.5887 slope=1.5625 100%|██████████| 4/4 [04:38<00:00, 69.86s/it] 100%|██████████| 4/4 [04:38<00:00, 69.69s/it] [INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Prompt executed in 443.20 seconds [INFO] got prompt [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. [H3TURBO sampler] sigmas=[1.0, 0.9837, 0.9601, 0.9231, 0.8575, 0.7064, 0.0] x.shape=(1, 1, 3423168) dtype=torch.float32 v_numel=3409920 a_numel=13248 shapes=[torch.Size([1, 24, 37, 80, 48]), torch.Size([1, 32, 2, 207])] 0%| | 0/6 [00:00 lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=1000.00 video_rms=0.9999 audio_rms=1.0110 dtype=torch.bfloat16 [H3TURBO step 0] sv=1.0000->0.9837 denoised_rms=2.0425 video: x_rms=0.9844 v_rms=2.2590 audio: x_rms=0.9470 v_rms=4.7463 slope=4.0000 17%|█▋ | 1/6 [01:09<05:47, 69.47s/it][H3TURBO fwd pruned/bypass] call#6 qkv_proj.forward_owner=Linear (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=983.68 video_rms=0.9844 audio_rms=0.9470 dtype=torch.bfloat16 [H3TURBO step 1] sv=0.9837->0.9601 denoised_rms=1.1436 video: x_rms=0.9617 v_rms=1.5078 audio: x_rms=0.8681 v_rms=3.9842 slope=3.6354 33%|███▎ | 2/6 [02:18<04:36, 69.14s/it][H3TURBO step 2] sv=0.9601->0.9231 denoised_rms=1.1721 video: x_rms=0.9278 v_rms=1.5356 audio: x_rms=0.7668 v_rms=3.4380 slope=3.1898 50%|█████ | 3/6 [03:28<03:28, 69.47s/it][H3TURBO step 3] sv=0.9231->0.8575 denoised_rms=1.1144 video: x_rms=0.8733 v_rms=1.4876 audio: x_rms=0.6361 v_rms=2.8796 slope=2.6406 67%|██████▋ | 4/6 [04:38<02:19, 69.80s/it][H3TURBO step 4] sv=0.8575->0.7064 denoised_rms=1.0925 video: x_rms=0.7798 v_rms=1.4677 audio: x_rms=0.4830 v_rms=2.0929 slope=1.9630 83%|████████▎ | 5/6 [05:48<01:09, 69.96s/it][H3TURBO step 5] sv=0.7064->0.0000 denoised_rms=1.0916 video: x_rms=1.0928 v_rms=1.4610 audio: x_rms=0.4470 v_rms=1.1083 slope=1.1307 100%|██████████| 6/6 [06:59<00:00, 70.06s/it] 100%|██████████| 6/6 [06:59<00:00, 69.84s/it] [INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Prompt executed in 581.45 seconds [INFO] got prompt [MiniMaxH3TurboLoRA] pruned base [bypass]: lora=minimax_h3_turbo_4step_ema_ckpt850.safetensors strength=1.0 | 208 backbone modules, 208 adapters bound, 1 bypass injections set + 51 adaln injected at run time | model=MiniMaxH3 weight_dtype=torch.bfloat16 weight_dev=cpu [INFO] Applying MiniMax H3 Memory Efficient Sage Attention Patch to all transformer blocks [INFO] Requested to load MiniMaxH3 [INFO] 0 models unloaded. [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. [H3TURBO sampler] sigmas=[1.0, 0.973, 0.9231, 0.8, 0.0] x.shape=(1, 1, 3423168) dtype=torch.float32 v_numel=3409920 a_numel=13248 shapes=[torch.Size([1, 24, 37, 80, 48]), torch.Size([1, 32, 2, 207])] 0%| | 0/4 [00:00 lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=1000.00 video_rms=0.9999 audio_rms=1.0110 dtype=torch.bfloat16 [H3TURBO step 0] sv=1.0000->0.9730 denoised_rms=1.5753 video: x_rms=0.9746 v_rms=1.8323 audio: x_rms=0.9117 v_rms=5.1878 slope=4.0000 25%|██▌ | 1/4 [01:17<03:53, 77.97s/it][H3TURBO fwd pruned/bypass] call#2 qkv_proj.forward_owner=BypassForwardHook (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=972.97 video_rms=0.9746 audio_rms=0.9117 dtype=torch.bfloat16 [H3TURBO step 1] sv=0.9730->0.9231 denoised_rms=1.2000 video: x_rms=0.9278 v_rms=1.5592 audio: x_rms=0.7773 v_rms=4.3009 slope=3.4225 50%|█████ | 2/4 [02:33<02:32, 76.34s/it][H3TURBO fwd pruned/bypass] call#3 qkv_proj.forward_owner=BypassForwardHook (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=923.08 video_rms=0.9278 audio_rms=0.7773 dtype=torch.bfloat16 [H3TURBO step 2] sv=0.9231->0.8000 denoised_rms=1.0783 video: x_rms=0.8277 v_rms=1.4710 audio: x_rms=0.5939 v_rms=3.3587 slope=2.6406 75%|███████▌ | 3/4 [03:48<01:15, 75.82s/it][H3TURBO fwd pruned/bypass] call#4 qkv_proj.forward_owner=BypassForwardHook (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=800.00 video_rms=0.8277 audio_rms=0.5939 dtype=torch.bfloat16 [H3TURBO step 3] sv=0.8000->0.0000 denoised_rms=1.0356 video: x_rms=1.0355 v_rms=1.4381 audio: x_rms=0.5139 v_rms=1.6426 slope=1.5625 100%|██████████| 4/4 [05:03<00:00, 75.61s/it] 100%|██████████| 4/4 [05:03<00:00, 75.91s/it] [INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Prompt executed in 480.97 seconds [INFO] got prompt [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. [H3TURBO sampler] sigmas=[1.0, 0.9837, 0.9601, 0.9231, 0.8575, 0.7064, 0.0] x.shape=(1, 1, 3423168) dtype=torch.float32 v_numel=3409920 a_numel=13248 shapes=[torch.Size([1, 24, 37, 80, 48]), torch.Size([1, 32, 2, 207])] 0%| | 0/6 [00:00 lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=1000.00 video_rms=0.9999 audio_rms=1.0110 dtype=torch.bfloat16 [H3TURBO step 0] sv=1.0000->0.9837 denoised_rms=1.5753 video: x_rms=0.9844 v_rms=1.8323 audio: x_rms=0.9483 v_rms=5.1878 slope=4.0000 17%|█▋ | 1/6 [01:16<06:20, 76.08s/it][H3TURBO fwd pruned/bypass] call#6 qkv_proj.forward_owner=BypassForwardHook (BypassForwardHook => lora ACTIVE; else => BASE ONLY!) is_injected=False timestep=983.68 video_rms=0.9844 audio_rms=0.9483 dtype=torch.bfloat16 [H3TURBO step 1] sv=0.9837->0.9601 denoised_rms=1.2198 video: x_rms=0.9615 v_rms=1.5697 audio: x_rms=0.8720 v_rms=4.4290 slope=3.6354 33%|███▎ | 2/6 [02:31<05:01, 75.49s/it][H3TURBO step 2] sv=0.9601->0.9231 denoised_rms=1.1129 video: x_rms=0.9269 v_rms=1.4927 audio: x_rms=0.7764 v_rms=4.0933 slope=3.1898 50%|█████ | 3/6 [03:46<03:45, 75.30s/it][H3TURBO step 3] sv=0.9231->0.8575 denoised_rms=1.0620 video: x_rms=0.8701 v_rms=1.4586 audio: x_rms=0.6569 v_rms=3.3132 slope=2.6406 67%|██████▋ | 4/6 [05:01<02:30, 75.26s/it][H3TURBO step 4] sv=0.8575->0.7064 denoised_rms=1.0332 video: x_rms=0.7680 v_rms=1.4389 audio: x_rms=0.5227 v_rms=2.1977 slope=1.9630 83%|████████▎ | 5/6 [06:16<01:15, 75.20s/it][H3TURBO step 5] sv=0.7064->0.0000 denoised_rms=1.0187 video: x_rms=1.0195 v_rms=1.4237 audio: x_rms=0.5092 v_rms=1.1555 slope=1.1307 100%|██████████| 6/6 [07:31<00:00, 75.15s/it] 100%|██████████| 6/6 [07:31<00:00, 75.27s/it] [INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Prompt executed in 00:10:26 [INFO] got prompt [INFO] EasyCache enabled - threshold: 0.2, start_percent: 0.15, end_percent: 0.95 [INFO] Requested to load MiniMaxH3 [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB. [H3TURBO sampler] sigmas=[1.0, 0.973, 0.9231, 0.8, 0.0] x.shape=(1, 1, 3423168) dtype=torch.float32 v_numel=3409920 a_numel=13248 shapes=[torch.Size([1, 24, 37, 80, 48]), torch.Size([1, 32, 2, 207])] 0%| | 0/4 [00:000.9730 denoised_rms=1.5753 video: x_rms=0.9746 v_rms=1.8323 audio: x_rms=0.9117 v_rms=5.1878 slope=4.0000 25%|██▌ | 1/4 [01:15<03:47, 75.86s/it][INFO] EasyCache [verbose] - x_prev_subsampled: torch.Size([1, 24, 37, 10, 6]) [H3TURBO step 1] sv=0.9730->0.9231 denoised_rms=1.2000 video: x_rms=0.9278 v_rms=1.5592 audio: x_rms=0.7773 v_rms=4.3009 slope=3.4225 50%|█████ | 2/4 [02:31<02:31, 75.53s/it][INFO] EasyCache [verbose] - output_change_rate: 0.43359375 [INFO] EasyCache [verbose] - x_prev_subsampled: torch.Size([1, 24, 37, 10, 6]) [H3TURBO step 2] sv=0.9231->0.8000 denoised_rms=1.0783 video: x_rms=0.8277 v_rms=1.4710 audio: x_rms=0.5939 v_rms=3.3587 slope=2.6406 75%|███████▌ | 3/4 [03:46<01:15, 75.66s/it][INFO] EasyCache [verbose] - NOT skipping step; cumulative_change_rate: 1.078125, reuse_threshold: 0.2 [INFO] EasyCache [verbose] - approx_output_change_rate: 1.078125 [INFO] EasyCache [verbose] - output_change_rate: 0.25 [INFO] EasyCache [verbose] - x_prev_subsampled: torch.Size([1, 24, 37, 10, 6]) [H3TURBO step 3] sv=0.8000->0.0000 denoised_rms=1.0356 video: x_rms=1.0355 v_rms=1.4381 audio: x_rms=0.5139 v_rms=1.6426 slope=1.5625 100%|██████████| 4/4 [05:03<00:00, 75.82s/it] 100%|██████████| 4/4 [05:03<00:00, 75.76s/it] [INFO] EasyCache [verbose] - output_change_rates 2: [0.43359375, 0.25] [INFO] EasyCache [verbose] - approx_output_change_rates 1: [1.078125] [INFO] EasyCache - skipped 0/4 steps (1.00x speedup). [INFO] Model MiniMaxH3AudioVAE prepared for dynamic VRAM loading. 576MB Staged. 0 patches attached. Force pre-loaded 401 weights: 539 KB. [INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB. [INFO] Prompt executed in 458.86 seconds