Q8_CR and LoRA - Compatibility Inquiry

by psdboi - opened 6 days ago

Are LoRAs unable to be applied using the Q8_CR? I tried it and it would not work. However, I tried the Q8_0, and that one worked just fine; it generated with LoRA.

molbal

Owner 6 days ago

I haven't had any issues but I'll do a few tests to verify it works.

Do you get an error, or just an image generated seemingly without the LoRA?

Theoretically, Q8_QR should behave identically to an INT8 safetensors checkpoint

psdboi

6 days ago

•

edited 6 days ago

Hi! No errors. I've tested generation with the LoRA Loader node enabled and then bypassed, and the result is the same behavior as if the image were generated without the LoRA applied. I used the official Krea demo LoRAs and the provided prompts for testing. Thanks for looking into it.

molbal

Owner 5 days ago

Could you share your workflow please? I cannot reproduce it, I tried it both on my windows and linux.

In this example (should have the comfyui workflow in metadata) I used a shadowheart character and pixar style LoRA and it seems to apply:

Without it the output is different:

I'm not ruling out that there is a bug somewhere, but I need some additional help to find it :)

psdboi

5 days ago

•

edited 5 days ago

Sure, I can do that. I'm not sure if it helps but I'm on ComfyUI v0.28.0 (workflow specific reasons), I used Comfy-Org/Krea-2: krea2_turbo_bf16.safetensors and your conversion tool to make the Q8_CR and Q8_0 .gguf used in my workflow, and krea/Krea-2-LoRA-sunsetblur if you want to reproduce using the exact LoRA.

Edit: It's worth mentioning that I installed the tools in a standalone directory, separate of the ComfyUI custom node installation. The only error the came up was when I was making the Q8_CR. The program was looking for 'comfy_kitchen,' and since it was outside of ComfyUI directory, I just manually installed it to the standalone to proceed.

Krea2 Turbo Q8_0 with LoRA *ON:

Krea2 Turbo Q8_0 with LoRA *OFF:

Krea2 Turbo Q8_CR with LoRA *ON:

Krea2 Turbo Q8_CR with LoRA *OFF:

As you see, Q8_0 applies the style LoRA correctly, while Q8_CR fails at this; only slightly shifting the original around.

molbal

Owner 5 days ago

Thanks for the workflows, I think I found the problem!

It is the Unet Loader (GGUF, Legacy node) loader. If you swap it to the Unet Loader (GGUF, Dynamic VRAM) it works. I could reproduce it this way.

Once you verify this resolves the problem on your machine too, I'll add a note to the docs

psdboi

5 days ago

I tried your suggestion. Before I could get a result, the workflow threw an error from the 'Unet Loader (GGUF, Dynamic VRAM),' unfortunately.

For your review, the error log is as follows:

Node threw an error during execution.

ComfyUI Error Report

Error Details

Node ID: 236
Node Type: UnetLoaderGGUFDynamicVRAM
Exception Type: RuntimeError
Exception Message: RuntimeError: Dynamic VRAM is not enabled in this ComfyUI installation. Start ComfyUI without --disable-dynamic-vram and use a build that supports DynamicVRAM.

Stack Trace

  File "D:\Pinokio\api\comfy.git\app\execution.py", line 543, in execute
    output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

  File "D:\Pinokio\api\comfy.git\app\execution.py", line 342, in get_output_data
    return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

  File "D:\Pinokio\api\comfy.git\app\execution.py", line 316, in _async_map_node_over_list
    await process_inputs(input_dict, i)

  File "D:\Pinokio\api\comfy.git\app\execution.py", line 304, in process_inputs
    result = f(**inputs)

  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 862, in load_unet
    return (_load_dynamic_gguf_unet(unet_path, progress=GGUFLoadProgress([unet_path])),)

  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 778, in _load_dynamic_gguf_unet
    _require_dynamic_vram()

  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 748, in _require_dynamic_vram
    raise RuntimeError(

System Information

ComfyUI Version: 0.28.0
Arguments: main.py
OS: win32
Python Version: 3.10.20 | packaged by conda-forge | (main, Jun 11 2026, 03:28:51) [MSC v.1944 64 bit (AMD64)]
Embedded Python: false
PyTorch Version: 2.7.0+cu128

Devices

Name: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
- Type: cuda
- VRAM Total: 25769279488
- VRAM Free: 24429592576
- Torch VRAM Total: 33554432
- Torch VRAM Free: 25034752

Logs

2026-09-04T14:38:21.616510 - [32m[INFO][0m setup plugin alembic.autogenerate.schemas
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.tables
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.types
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.constraints
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.defaults
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.comments
2026-09-04T14:38:21.617510 - [32m[INFO][0m setup plugin alembic.autogenerate.checkconstraint_byname
2026-09-04T14:38:28.408799 - [START] Security scan2026-09-04T14:38:28.408799 - 
2026-09-04T14:38:39.711408 - [DONE] Security scan2026-09-04T14:38:39.711408 - 
2026-09-04T14:38:41.033180 - ## ComfyUI-Manager: installing dependencies done.2026-09-04T14:38:41.033180 - 
2026-09-04T14:38:41.033180 - ** ComfyUI startup time:2026-09-04T14:38:41.033180 -  2026-09-04T14:38:41.033180 - 2026-09-04 14:38:41.0332026-09-04T14:38:41.033180 - 
2026-09-04T14:38:41.033180 - ** Platform:2026-09-04T14:38:41.033180 -  2026-09-04T14:38:41.034219 - Windows2026-09-04T14:38:41.034219 - 
2026-09-04T14:38:41.034219 - ** Python version:2026-09-04T14:38:41.034219 -  2026-09-04T14:38:41.034219 - 3.10.20 | packaged by conda-forge | (main, Jun 11 2026, 03:28:51) [MSC v.1944 64 bit (AMD64)]2026-09-04T14:38:41.034219 - 
2026-09-04T14:38:41.034219 - ** Python executable:2026-09-04T14:38:41.034219 -  2026-09-04T14:38:41.034219 - D:\Pinokio\api\comfy.git\app\env\Scripts\python.exe2026-09-04T14:38:41.034219 - 
2026-09-04T14:38:41.034219 - ** ComfyUI Path:2026-09-04T14:38:41.034219 -  2026-09-04T14:38:41.034219 - D:\Pinokio\api\comfy.git\app2026-09-04T14:38:41.035210 - 
2026-09-04T14:38:41.035210 - ** ComfyUI Base Folder Path:2026-09-04T14:38:41.035210 -  2026-09-04T14:38:41.035210 - D:\Pinokio\api\comfy.git\app2026-09-04T14:38:41.035210 - 
2026-09-04T14:38:41.035210 - ** User directory:2026-09-04T14:38:41.035210 -  2026-09-04T14:38:41.035210 - D:\Pinokio\api\comfy.git\app\user2026-09-04T14:38:41.035210 - 
2026-09-04T14:38:41.035210 - ** ComfyUI-Manager config path:2026-09-04T14:38:41.035210 -  2026-09-04T14:38:41.036210 - D:\Pinokio\api\comfy.git\app\user\__manager\config.ini2026-09-04T14:38:41.036210 - 
2026-09-04T14:38:41.036210 - ** Log path:2026-09-04T14:38:41.036210 -  2026-09-04T14:38:41.036210 - D:\Pinokio\api\comfy.git\app\user\comfyui.log2026-09-04T14:38:41.036210 - 
2026-09-04T14:38:45.449998 - [32m[INFO][0m 
Prestartup times for custom nodes:
2026-09-04T14:38:45.449998 - [32m[INFO][0m    0.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\rgthree-comfy
2026-09-04T14:38:45.450997 - [32m[INFO][0m    0.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-easy-use
2026-09-04T14:38:45.450997 - [32m[INFO][0m   22.7 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Manager
2026-09-04T14:38:45.450997 - [32m[INFO][0m 
2026-09-04T14:39:00.478209 - D:\Pinokio\api\comfy.git\app\env\lib\site-packages\torch\cuda\__init__.py:61: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
  import pynvml  # type: ignore[import]
2026-09-04T14:39:22.897430 - [1m[33m[WARNING][0m WARNING: You need pytorch with cu130 or higher to use optimized CUDA operations.
2026-09-04T14:39:22.897430 - [32m[INFO][0m Found comfy_kitchen backend cuda: {'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_', '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', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', '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', 'quantize_w4a8_int8_weight', '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_', 'rotate_int8_convrot_weight', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
2026-09-04T14:39:22.898440 - [32m[INFO][0m Found comfy_kitchen backend hip: {'available': False, 'disabled': False, 'unavailable_reason': 'PyTorch ROCm/HIP runtime not available', 'capabilities': []}
2026-09-04T14:39:22.898440 - [32m[INFO][0m 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', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', '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', 'quantize_w4a8_int8_weight', '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_', 'rotate_int8_convrot_weight', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
2026-09-04T14:39:22.898440 - [32m[INFO][0m 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', 'na3d', '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_', 'w4a8_int8_linear']}
2026-09-04T14:39:22.979510 - [32m[INFO][0m Checkpoint files will always be loaded safely.
2026-09-04T14:39:23.318468 - [32m[INFO][0m Total VRAM 24576 MB, total RAM 130867 MB
2026-09-04T14:39:23.319466 - [32m[INFO][0m pytorch version: 2.7.0+cu128
2026-09-04T14:39:23.321006 - [32m[INFO][0m Set vram state to: NORMAL_VRAM
2026-09-04T14:39:23.321006 - [32m[INFO][0m Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
2026-09-04T14:39:23.343720 - [32m[INFO][0m Using async weight offloading with 2 streams
2026-09-04T14:39:23.343720 - [32m[INFO][0m Enabled pinned memory 52346.0
2026-09-04T14:39:29.017297 - [32m[INFO][0m Using pytorch attention
2026-09-04T14:40:13.094653 - [1m[33m[WARNING][0m Unsupported Pytorch detected. DynamicVRAM support requires Pytorch version 2.8 or later. Falling back to legacy ModelPatcher. VRAM estimates may be unreliable especially on Windows
2026-09-04T14:40:13.094653 - [32m[INFO][0m Python version: 3.10.20 | packaged by conda-forge | (main, Jun 11 2026, 03:28:51) [MSC v.1944 64 bit (AMD64)]
2026-09-04T14:40:13.094653 - [32m[INFO][0m ComfyUI version: 0.28.0
2026-09-04T14:40:13.097654 - [32m[INFO][0m comfy-aimdo version: 0.4.13
2026-09-04T14:40:13.099667 - [32m[INFO][0m comfy-kitchen version: 0.2.31
2026-09-04T14:40:13.142896 - [32m[INFO][0m comfyui-frontend-package version: 1.49.6
2026-09-04T14:40:13.143899 - [32m[INFO][0m comfyui-workflow-templates version: 0.11.44
2026-09-04T14:40:13.143899 - [32m[INFO][0m comfyui-embedded-docs version: 0.5.10
2026-09-04T14:40:13.143899 - [32m[INFO][0m comfy-kitchen version: 0.2.31
2026-09-04T14:40:13.143899 - [32m[INFO][0m comfy-aimdo version: 0.4.13
2026-09-04T14:40:13.148866 - [32m[INFO][0m [Prompt Server] web root: D:\Pinokio\api\comfy.git\app\env\lib\site-packages\comfyui_frontend_package\static
2026-09-04T14:40:13.149869 - [32m[INFO][0m Asset seeder disabled
2026-09-04T14:40:21.976538 - [32m[INFO][0m No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
2026-09-04T14:40:28.417891 - [Crystools [0;32mINFO[0m] Crystools version: 1.27.4
2026-09-04T14:40:28.688464 - [Crystools [0;32mINFO[0m] Platform release: 10
2026-09-04T14:40:28.689464 - [Crystools [0;32mINFO[0m] JETSON: Not detected.
2026-09-04T14:40:28.691973 - [Crystools [0;32mINFO[0m] CPU: Intel(R) Core(TM) i9-10900KF CPU @ 3.70GHz - Arch: AMD64 - OS: Windows 10
2026-09-04T14:40:28.702486 - [Crystools [0;32mINFO[0m] pynvml (NVIDIA) initialized.
2026-09-04T14:40:28.703490 - [Crystools [0;32mINFO[0m] GPU/s:
2026-09-04T14:40:28.721004 - [Crystools [0;32mINFO[0m] 0) NVIDIA GeForce RTX 3090
2026-09-04T14:40:28.721004 - [Crystools [0;32mINFO[0m] NVIDIA Driver: 610.88
2026-09-04T14:40:42.404256 - [ComfyUI-Easy-Use] server: 2026-09-04T14:40:42.404256 - v1.3.6 2026-09-04T14:40:42.404256 - Loaded2026-09-04T14:40:42.404256 - 
2026-09-04T14:40:42.404256 - [ComfyUI-Easy-Use] web root: 2026-09-04T14:40:42.404256 - D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-easy-use\web_version/v2 2026-09-04T14:40:42.404256 - Loaded2026-09-04T14:40:42.405255 - 
2026-09-04T14:40:44.667963 - [32m[INFO][0m ComfyUI-GGUF: Partial torch compile only, consider updating pytorch
2026-09-04T14:40:44.798606 - [32m[INFO][0m ### Loading: ComfyUI-Impact-Pack (V8.28.3)
2026-09-04T14:40:46.365850 - [32m[INFO][0m [Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.
2026-09-04T14:40:46.379044 - [32m[INFO][0m [Impact Pack] Wildcards loading done.
2026-09-04T14:40:46.471858 - [32m[INFO][0m ### Loading: ComfyUI-Impact-Subpack (V1.3.5)
2026-09-04T14:40:46.530853 - [32m[INFO][0m [Impact Pack/Subpack] Using folder_paths to determine whitelist path: D:\Pinokio\api\comfy.git\app\user\default\ComfyUI-Impact-Subpack\model-whitelist.txt
2026-09-04T14:40:46.532195 - [32m[INFO][0m [Impact Pack/Subpack] Ensured whitelist directory exists: D:\Pinokio\api\comfy.git\app\user\default\ComfyUI-Impact-Subpack
2026-09-04T14:40:46.538195 - [32m[INFO][0m [Impact Pack/Subpack] Loaded 0 model(s) from whitelist: D:\Pinokio\api\comfy.git\app\user\default\ComfyUI-Impact-Subpack\model-whitelist.txt
2026-09-04T14:40:47.656929 - [32m[INFO][0m [Impact Subpack] ultralytics_bbox: D:\Pinokio\api\comfy.git\app\models\ultralytics\bbox
2026-09-04T14:40:47.656929 - [32m[INFO][0m [Impact Subpack] ultralytics_segm: D:\Pinokio\api\comfy.git\app\models\ultralytics\segm
2026-09-04T14:40:47.929060 - [1m[33m[WARNING][0m KJNodes: MiniMax nodes could not be imported. MiniMax nodes will be unavailable. Error: No module named 'comfy.ldm.minimax.model'
Traceback (most recent call last):
  File "D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-kjnodes\__init__.py", line 373, in <module>
    from .nodes.minimax_nodes import MiniMaxChunkFeedForward, MiniMaxLowVRAMAttention
  File "D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-kjnodes\nodes\minimax_nodes.py", line 10, in <module>
    from comfy.ldm.minimax.model import _mod_scale_shift, _mod_gate
ModuleNotFoundError: No module named 'comfy.ldm.minimax.model'
2026-09-04T14:40:48.307840 - [32m[INFO][0m ### Loading: ComfyUI-Manager (V3.41)
2026-09-04T14:40:48.307840 - [32m[INFO][0m [ComfyUI-Manager] network_mode: public
2026-09-04T14:40:48.308841 - [32m[INFO][0m [ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
2026-09-04T14:40:48.451990 - [32m[INFO][0m ### ComfyUI Revision: 5594 [700821e1] *DETACHED | Released on '2026-07-15'
2026-09-04T14:40:49.510827 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
2026-09-04T14:40:49.748204 - --- Logging error ---
2026-09-04T14:40:49.791334 - Traceback (most recent call last):
2026-09-04T14:40:49.793337 -   File "D:\Pinokio\bin\miniforge\lib\logging\__init__.py", line 1103, in emit
    stream.write(msg + self.terminator)
2026-09-04T14:40:49.796337 -   File "D:\Pinokio\api\comfy.git\app\app\logger.py", line 66, in write
    super().write(data)
2026-09-04T14:40:49.799338 - OSError: [WinError 1] Incorrect function
2026-09-04T14:40:49.801853 - Call stack:
2026-09-04T14:40:49.805853 -   File "D:\Pinokio\bin\miniforge\lib\threading.py", line 973, in _bootstrap
    self._bootstrap_inner()
2026-09-04T14:40:49.807852 -   File "D:\Pinokio\bin\miniforge\lib\threading.py", line 1016, in _bootstrap_inner
    self.run()
2026-09-04T14:40:49.814357 -   File "D:\Pinokio\bin\miniforge\lib\threading.py", line 953, in run
    self._target(*self._args, **self._kwargs)
2026-09-04T14:40:49.816358 -   File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Manager\glob\manager_server.py", line 2094, in <lambda>
    threading.Thread(target=lambda: asyncio.run(default_cache_update())).start()
2026-09-04T14:40:49.819359 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\runners.py", line 44, in run
    return loop.run_until_complete(main)
2026-09-04T14:40:49.820861 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\base_events.py", line 636, in run_until_complete
    self.run_forever()
2026-09-04T14:40:49.822872 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\windows_events.py", line 321, in run_forever
    super().run_forever()
2026-09-04T14:40:49.824871 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\base_events.py", line 603, in run_forever
    self._run_once()
2026-09-04T14:40:49.826870 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\base_events.py", line 1909, in _run_once
    handle._run()
2026-09-04T14:40:49.829871 -   File "D:\Pinokio\bin\miniforge\lib\asyncio\events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
2026-09-04T14:40:49.831377 -   File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Manager\glob\manager_server.py", line 2070, in get_cache
    logging.info(f"[ComfyUI-Manager] default cache updated: {uri}")
2026-09-04T14:40:49.833384 - Message: '[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json'
Arguments: ()
2026-09-04T14:40:49.875931 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
2026-09-04T14:40:49.949213 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
2026-09-04T14:40:50.085829 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
2026-09-04T14:40:50.213831 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
2026-09-04T14:40:53.732238 - [D:\Pinokio\api\comfy_x_git\app\custom_nodes\comfyui_controlnet_aux] | INFO -> Using ckpts path: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_controlnet_aux\ckpts2026-09-04T14:40:53.734213 - 
2026-09-04T14:40:53.736238 - [D:\Pinokio\api\comfy_x_git\app\custom_nodes\comfyui_controlnet_aux] | INFO -> Using symlinks: False2026-09-04T14:40:53.737239 - 
2026-09-04T14:40:53.739241 - [D:\Pinokio\api\comfy_x_git\app\custom_nodes\comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider']2026-09-04T14:40:53.741781 - 
2026-09-04T14:40:54.081345 - D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_controlnet_aux\node_wrappers\dwpose.py:26: UserWarning: DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly
  warnings.warn("DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly")
2026-09-04T14:40:54.573210 - # 😺dzNodes: LayerStyle -> 2026-09-04T14:40:54.573210 - Cannot import name 'guidedFilter' from 'cv2.ximgproc'
A few nodes cannot works properly, while most nodes are not affected. Please REINSTALL package 'opencv-contrib-python'.
For detail refer to 2026-09-04T14:40:54.573210 - https://github.com/chflame163/ComfyUI_LayerStyle/issues/52026-09-04T14:40:54.573210 - 
2026-09-04T14:40:55.907415 - USDU batch patches applied successfully.
2026-09-04T14:40:55.907415 - [32m[INFO][0m USDU batch patches applied successfully.
2026-09-04T14:40:57.105059 - 
2026-09-04T14:40:57.105059 - [rgthree-comfy] Loaded 48 exciting nodes. 🎉2026-09-04T14:40:57.105059 - 
2026-09-04T14:40:57.106060 - 
2026-09-04T14:40:57.106060 - [rgthree-comfy] ComfyUI's new Node 2.0 rendering may be incompatible with some rgthree-comfy nodes and features, breaking some rendering as well as losing the ability to access a node's properties (a vital part of many nodes). It also appears to run MUCH more slowly spiking CPU usage and causing jankiness and unresponsiveness, especially with large workflows. Personally I am not planning to use the new Nodes 2.0 and, unfortunately, am not able to invest the time to investigate and overhaul rgthree-comfy where needed. If you have issues when Nodes 2.0 is enabled, I'd urge you to switch it off as well and join me in hoping ComfyUI is not planning to deprecate the existing, stable canvas rendering all together.
2026-09-04T14:40:57.106060 - 
2026-09-04T14:40:57.203777 - ⚡ SeedVR2 optimizations check: SageAttention ✅ | Flash Attention ✅ | Triton ✅2026-09-04T14:40:57.203777 - 
2026-09-04T14:40:58.471743 - 📊 Initial CUDA memory: 22.73GB free / 24.00GB total2026-09-04T14:40:58.471743 - 
2026-09-04T14:40:59.468984 - [32m[INFO][0m 
Import times for custom nodes:
2026-09-04T14:40:59.468984 - [32m[INFO][0m    0.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\websocket_image_save.py
2026-09-04T14:40:59.470013 - [32m[INFO][0m    0.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-SelectStringFromListWithIndex
2026-09-04T14:40:59.470013 - [32m[INFO][0m    0.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-inpaint-cropandstitch
2026-09-04T14:40:59.470013 - [32m[INFO][0m    0.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_stringessentials
2026-09-04T14:40:59.470013 - [32m[INFO][0m    0.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-custom-scripts
2026-09-04T14:40:59.470013 - [32m[INFO][0m    0.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Krea2T-Enhancer
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-frame-interpolation
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.2 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_ultimatesdupscale
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.3 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.3 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Video-Crop
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.3 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-kjnodes
2026-09-04T14:40:59.471012 - [32m[INFO][0m    0.7 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_controlnet_aux
2026-09-04T14:40:59.472179 - [32m[INFO][0m    0.9 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-reference-latent-plus
2026-09-04T14:40:59.472179 - [32m[INFO][0m    1.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Manager
2026-09-04T14:40:59.472179 - [32m[INFO][0m    1.2 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\rgthree-comfy
2026-09-04T14:40:59.472179 - [32m[INFO][0m    1.2 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-impact-subpack
2026-09-04T14:40:59.472179 - [32m[INFO][0m    1.4 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui_layerstyle
2026-09-04T14:40:59.473177 - [32m[INFO][0m    1.6 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-impact-pack
2026-09-04T14:40:59.473177 - [32m[INFO][0m    1.9 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-florence2
2026-09-04T14:40:59.473177 - [32m[INFO][0m    2.3 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\seedvr2_videoupscaler
2026-09-04T14:40:59.473177 - [32m[INFO][0m    3.0 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-Crystools
2026-09-04T14:40:59.473177 - [32m[INFO][0m    3.1 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-videohelpersuite
2026-09-04T14:40:59.474178 - [32m[INFO][0m   13.6 seconds: D:\Pinokio\api\comfy.git\app\custom_nodes\comfyui-easy-use
2026-09-04T14:40:59.474178 - [32m[INFO][0m 
2026-09-04T14:40:59.548090 - [32m[INFO][0m Context impl SQLiteImpl.
2026-09-04T14:40:59.548090 - [32m[INFO][0m Will assume non-transactional DDL.
2026-09-04T14:41:01.789444 - [32m[INFO][0m Using RAM pressure cache.
2026-09-04T14:41:01.790950 - [32m[INFO][0m Starting server

2026-09-04T14:41:01.790950 - [32m[INFO][0m To see the GUI go to: http://127.0.0.1:8188
2026-09-04T14:41:12.827176 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/buttonGroup.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:41:12.838685 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:41:12.924284 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/clipspace.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:41:12.928283 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/groupNode.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:41:12.944360 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/widgetInputs.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:42:52.578970 - FETCH ComfyRegistry Data [DONE]2026-09-04T14:42:52.578970 - 
2026-09-04T14:42:53.121279 - [32m[INFO][0m [ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
2026-09-04T14:42:53.210946 - FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json2026-09-04T14:42:53.367729 -  [DONE]2026-09-04T14:42:53.367729 - 
2026-09-04T14:42:53.552695 - [32m[INFO][0m [ComfyUI-Manager] All startup tasks have been completed.
2026-09-04T14:48:24.329816 - [1m[33m[WARNING][0m [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/button.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
2026-09-04T14:49:04.431026 - [32m[INFO][0m got prompt
2026-09-04T14:49:04.852927 - [32m[INFO][0m Using pytorch attention in VAE
2026-09-04T14:49:04.854929 - [32m[INFO][0m Using pytorch attention in VAE
2026-09-04T14:49:09.613409 - [32m[INFO][0m VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
2026-09-04T14:54:05.013043 - [32m[INFO][0m Requested to load Krea2TEModel_
2026-09-04T14:54:05.108639 - [32m[INFO][0m loaded completely;  8464.46 MB loaded, full load: True
2026-09-04T14:54:05.117302 - [32m[INFO][0m CLIP/text encoder model load device: cpu, offload device: cpu, current: cpu, dtype: torch.float16
2026-09-04T14:54:06.216574 - [1m[31m[ERROR][0m !!! Exception during processing !!! Dynamic VRAM is not enabled in this ComfyUI installation. Start ComfyUI without --disable-dynamic-vram and use a build that supports DynamicVRAM.
2026-09-04T14:54:06.265069 - [1m[31m[ERROR][0m Traceback (most recent call last):
  File "D:\Pinokio\api\comfy.git\app\execution.py", line 543, in execute
    output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
  File "D:\Pinokio\api\comfy.git\app\execution.py", line 342, in get_output_data
    return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
  File "D:\Pinokio\api\comfy.git\app\execution.py", line 316, in _async_map_node_over_list
    await process_inputs(input_dict, i)
  File "D:\Pinokio\api\comfy.git\app\execution.py", line 304, in process_inputs
    result = f(**inputs)
  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 862, in load_unet
    return (_load_dynamic_gguf_unet(unet_path, progress=GGUFLoadProgress([unet_path])),)
  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 778, in _load_dynamic_gguf_unet
    _require_dynamic_vram()
  File "D:\Pinokio\api\comfy.git\app\custom_nodes\ComfyUI-GGUF\nodes.py", line 748, in _require_dynamic_vram
    raise RuntimeError(
RuntimeError: Dynamic VRAM is not enabled in this ComfyUI installation. Start ComfyUI without --disable-dynamic-vram and use a build that supports DynamicVRAM.

2026-09-04T14:54:06.267031 - [32m[INFO][0m [32mPrompt executed in 301.83 seconds[0m

Attached Workflow

Please make sure that workflow does not contain any sensitive information such as API keys or passwords.

{"id":"12582b72-3893-41d0-bbf8-ff40951e52a5","revision":0,"last_node_id":236,"last_link_id":258,"nodes":[{"id":60,"type":"UnetLoaderGGUF","pos":[-6100,-2370],"size":[400,70],"flags":{},"order":0,"mode":0,"inputs":[{"localized_name":"unet_name","name":"unet_name","type":"COMBO","widget":{"name":"unet_name"},"link":null}],"outputs":[{"localized_name":"MODEL","name":"MODEL","type":"MODEL","links":[]}],"properties":{"cnr_id":"comfyui-gguf-reboot","ver":"39cf309ca2c83381b007b3d0c9cfeaff8013a7a6","Node name for S&R":"UnetLoaderGGUF"},"widgets_values":["krea2_turbo_q8_cr.gguf"],"widgets_values_named":{"unet_name":"krea2_turbo_q8_cr.gguf"},"color":"#223","bgcolor":"#335"},{"id":76,"type":"ConditioningZeroOut","pos":[-4500,-2000],"size":[210,30],"flags":{},"order":9,"mode":0,"inputs":[{"localized_name":"conditioning","name":"conditioning","type":"CONDITIONING","link":45}],"outputs":[{"localized_name":"CONDITIONING","name":"CONDITIONING","type":"CONDITIONING","links":[249]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"ConditioningZeroOut"},"color":"#432","bgcolor":"#653"},{"id":72,"type":"EmptyLatentImage","pos":[-4200,-1910],"size":[270,110],"flags":{},"order":5,"mode":0,"inputs":[{"localized_name":"width","name":"width","type":"INT","widget":{"name":"width"},"link":243},{"localized_name":"height","name":"height","type":"INT","widget":{"name":"height"},"link":244},{"localized_name":"batch_size","name":"batch_size","type":"INT","widget":{"name":"batch_size"},"link":null}],"outputs":[{"localized_name":"LATENT","name":"LATENT","type":"LATENT","links":[41]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"EmptyLatentImage"},"widgets_values":[512,512,1],"widgets_values_named":{"width":512,"height":512,"batch_size":1},"color":"#323","bgcolor":"#535"},{"id":62,"type":"VAELoader","pos":[-3900,-1770],"size":[400,70],"flags":{},"order":1,"mode":0,"inputs":[{"localized_name":"vae_name","name":"vae_name","type":"COMBO","widget":{"name":"vae_name"},"link":null}],"outputs":[{"localized_name":"VAE","name":"VAE","type":"VAE","links":[252]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"VAELoader"},"widgets_values":["qwen_image_vae.safetensors"],"widgets_values_named":{"vae_name":"qwen_image_vae.safetensors"},"color":"#322","bgcolor":"#533"},{"id":235,"type":"PreviewImage","pos":[-3100,-2370],"size":[600,570],"flags":{},"order":13,"mode":0,"inputs":[{"localized_name":"images","name":"images","type":"IMAGE","link":253}],"outputs":[{"localized_name":"images","name":"images","type":"IMAGE","links":null}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"PreviewImage"}},{"id":196,"type":"VAEDecode","pos":[-3400,-2370],"size":[150,50],"flags":{},"order":11,"mode":0,"inputs":[{"localized_name":"samples","name":"samples","type":"LATENT","link":251},{"localized_name":"vae","name":"vae","type":"VAE","link":252}],"outputs":[{"localized_name":"IMAGE","name":"IMAGE","type":"IMAGE","links":[235,253,255]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"VAEDecode"},"color":"#323","bgcolor":"#535"},{"id":113,"type":"LayerUtility: PurgeVRAM","pos":[-3100,-1750],"size":[210,90],"flags":{"collapsed":true},"order":12,"mode":0,"inputs":[{"localized_name":"anything","name":"anything","type":"*","link":235},{"localized_name":"purge_cache","name":"purge_cache","type":"BOOLEAN","widget":{"name":"purge_cache"},"link":null},{"localized_name":"purge_models","name":"purge_models","type":"BOOLEAN","widget":{"name":"purge_models"},"link":null}],"outputs":[],"title":"Purge VRAM","properties":{"cnr_id":"comfyui_layerstyle","ver":"2.0.40","Node name for S&R":"LayerUtility: PurgeVRAM"},"widgets_values":[false,true],"widgets_values_named":{"purge_cache":false,"purge_models":true},"color":"rgba(38, 73, 116, 0.7)"},{"id":64,"type":"KSampler","pos":[-3800,-2370],"size":[270,270],"flags":{},"order":10,"mode":0,"inputs":[{"localized_name":"model","name":"model","type":"MODEL","link":246},{"localized_name":"positive","name":"positive","type":"CONDITIONING","link":250},{"localized_name":"negative","name":"negative","type":"CONDITIONING","link":249},{"localized_name":"latent_image","name":"latent_image","type":"LATENT","link":41},{"localized_name":"seed","name":"seed","type":"INT","widget":{"name":"seed"},"link":null},{"localized_name":"steps","name":"steps","type":"INT","widget":{"name":"steps"},"link":null},{"localized_name":"cfg","name":"cfg","type":"FLOAT","widget":{"name":"cfg"},"link":null},{"localized_name":"sampler_name","name":"sampler_name","type":"COMBO","widget":{"name":"sampler_name"},"link":null},{"localized_name":"scheduler","name":"scheduler","type":"COMBO","widget":{"name":"scheduler"},"link":null},{"localized_name":"denoise","name":"denoise","type":"FLOAT","widget":{"name":"denoise"},"link":null}],"outputs":[{"localized_name":"LATENT","name":"LATENT","type":"LATENT","links":[251]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"KSampler"},"widgets_values":[734292069461564,"fixed",8,1,"euler","simple",1],"widgets_values_named":{"seed":734292069461564,"control_after_generate":"fixed","steps":8,"cfg":1,"sampler_name":"euler","scheduler":"simple","denoise":1},"color":"#323","bgcolor":"#535"},{"id":119,"type":"ResolutionSelector","pos":[-4600,-1770],"size":[300,130],"flags":{},"order":2,"mode":0,"inputs":[{"localized_name":"aspect_ratio","name":"aspect_ratio","type":"COMBO","widget":{"name":"aspect_ratio"},"link":null},{"localized_name":"megapixels","name":"megapixels","type":"FLOAT","widget":{"name":"megapixels"},"link":null},{"localized_name":"multiple","name":"multiple","type":"INT","widget":{"name":"multiple"},"link":null}],"outputs":[{"localized_name":"width","name":"width","type":"INT","links":[243]},{"localized_name":"height","name":"height","type":"INT","links":[244]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"ResolutionSelector"},"widgets_values":["1:1 (Square)",1,8],"widgets_values_named":{"aspect_ratio":"1:1 (Square)","megapixels":1,"multiple":8},"color":"#323","bgcolor":"#535"},{"id":75,"type":"CLIPTextEncode","pos":[-5100,-2170],"size":[400,200],"flags":{},"order":8,"mode":0,"inputs":[{"localized_name":"clip","name":"clip","type":"CLIP","link":257},{"localized_name":"text","name":"text","type":"STRING","widget":{"name":"text"},"link":null}],"outputs":[{"localized_name":"CONDITIONING","name":"CONDITIONING","type":"CONDITIONING","links":[45,250]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"CLIPTextEncode"},"widgets_values":["A sorcerer in flowing midnight robes stands atop a cliff edge, arms raised as crackling energy gathers between his palms. The sky roils with energy, flowing down from the sky. ethereal motion blur style"],"widgets_values_named":{"text":"A sorcerer in flowing midnight robes stands atop a cliff edge, arms raised as crackling energy gathers between his palms. The sky roils with energy, flowing down from the sky. ethereal motion blur style"},"color":"#432","bgcolor":"#653"},{"id":131,"type":"ComfyUI-Krea2T-Enhancer","pos":[-5100,-2370],"size":[300,110],"flags":{},"order":7,"mode":0,"inputs":[{"localized_name":"model","name":"model","type":"MODEL","link":248},{"localized_name":"enabled","name":"enabled","type":"BOOLEAN","widget":{"name":"enabled"},"link":null},{"localized_name":"strength","name":"strength","type":"FLOAT","widget":{"name":"strength"},"link":null},{"localized_name":"debug","name":"debug","type":"BOOLEAN","widget":{"name":"debug"},"link":null}],"outputs":[{"localized_name":"model","name":"model","type":"MODEL","links":[246]}],"properties":{"cnr_id":"comfyui-krea2t-enhancer","ver":"1.2.0","Node name for S&R":"ComfyUI-Krea2T-Enhancer"},"widgets_values":[true,0.1,false],"widgets_values_named":{"enabled":true,"strength":0.1,"debug":false},"color":"#223","bgcolor":"#335"},{"id":234,"type":"LoraLoader","pos":[-5600,-2370],"size":[400,130],"flags":{},"order":6,"mode":0,"inputs":[{"localized_name":"model","name":"model","type":"MODEL","link":258},{"localized_name":"clip","name":"clip","type":"CLIP","link":256},{"localized_name":"lora_name","name":"lora_name","type":"COMBO","widget":{"name":"lora_name"},"link":null},{"localized_name":"strength_model","name":"strength_model","type":"FLOAT","widget":{"name":"strength_model"},"link":null},{"localized_name":"strength_clip","name":"strength_clip","type":"FLOAT","widget":{"name":"strength_clip"},"link":null}],"outputs":[{"localized_name":"MODEL","name":"MODEL","type":"MODEL","links":[248]},{"localized_name":"CLIP","name":"CLIP","type":"CLIP","links":[257]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"LoraLoader"},"widgets_values":["krea2_sunsetblur_style.safetensors",1,1],"widgets_values_named":{"lora_name":"krea2_sunsetblur_style.safetensors","strength_model":1,"strength_clip":1},"color":"#223","bgcolor":"#335"},{"id":162,"type":"CLIPLoader","pos":[-6100,-2170],"size":[400,110],"flags":{},"order":3,"mode":0,"inputs":[{"localized_name":"clip_name","name":"clip_name","type":"COMBO","widget":{"name":"clip_name"},"link":null},{"localized_name":"type","name":"type","type":"COMBO","widget":{"name":"type"},"link":null},{"localized_name":"device","name":"device","shape":7,"type":"COMBO","widget":{"name":"device"},"link":null}],"outputs":[{"localized_name":"CLIP","name":"CLIP","type":"CLIP","links":[256]}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"CLIPLoader"},"widgets_values":["qwen3vl_4b_bf16_instruct_abliterated.safetensors","krea2","cpu"],"widgets_values_named":{"clip_name":"qwen3vl_4b_bf16_instruct_abliterated.safetensors","type":"krea2","device":"cpu"},"color":"#432","bgcolor":"#653"},{"id":236,"type":"UnetLoaderGGUFDynamicVRAM","pos":[-6100,-2500],"size":[400,60],"flags":{},"order":4,"mode":0,"inputs":[{"localized_name":"unet_name","name":"unet_name","type":"COMBO","widget":{"name":"unet_name"},"link":null}],"outputs":[{"localized_name":"MODEL","name":"MODEL","type":"MODEL","links":[258]}],"properties":{"cnr_id":"comfyui-gguf-reboot","ver":"9f11f5aaeaf0ba662fe35d761c154fe340452193","Node name for S&R":"UnetLoaderGGUFDynamicVRAM"},"widgets_values":["krea2_turbo_q8_cr.gguf"],"widgets_values_named":{"unet_name":"krea2_turbo_q8_cr.gguf"}},{"id":123,"type":"SaveImage","pos":[-3100,-1710],"size":[210,60],"flags":{"collapsed":true},"order":14,"mode":4,"inputs":[{"localized_name":"images","name":"images","type":"IMAGE","link":255},{"localized_name":"filename_prefix","name":"filename_prefix","type":"STRING","widget":{"name":"filename_prefix"},"link":null}],"outputs":[{"localized_name":"images","name":"images","type":"IMAGE","links":null}],"properties":{"cnr_id":"comfy-core","ver":"0.28.0","Node name for S&R":"SaveImage"},"widgets_values":["ComfyUI"],"widgets_values_named":{"filename_prefix":"ComfyUI"}}],"links":[[41,72,0,64,3,"LATENT"],[45,75,0,76,0,"CONDITIONING"],[235,196,0,113,0,"IMAGE"],[243,119,0,72,0,"INT"],[244,119,1,72,1,"INT"],[246,131,0,64,0,"MODEL"],[248,234,0,131,0,"MODEL"],[249,76,0,64,2,"CONDITIONING"],[250,75,0,64,1,"CONDITIONING"],[251,64,0,196,0,"LATENT"],[252,62,0,196,1,"VAE"],[253,196,0,235,0,"IMAGE"],[255,196,0,123,0,"IMAGE"],[256,162,0,234,1,"CLIP"],[257,234,1,75,0,"CLIP"],[258,236,0,234,0,"MODEL"]],"groups":[],"config":{},"extra":{"ds":{"scale":0.19784466890013513,"offset":[7674.159106189465,3548.462277012378]},"anomalous_hashes":{},"frontendVersion":"1.49.6","VHS_latentpreview":false,"VHS_latentpreviewrate":0,"VHS_MetadataImage":true,"VHS_KeepIntermediate":true},"version":0.4}

Additional Context

(Please add any additional context or steps to reproduce the error here)

molbal

Owner 5 days ago

It's a good thing you pasted the log, looks like your pytorch version is old, see this line in it:

Unsupported Pytorch detected. DynamicVRAM support requires Pytorch version 2.8 or later

I do have PyTorch: 2.13.0+cu130 on my Windows laptop and PyTorch: 2.14.0.dev20260725+cu130 on my Ubuntu server, your installed PyTorch Version: 2.7.0+cu128 seems to be too old.

Are you in a position to attempt an update of Pytorch?

psdboi

5 days ago

Sure, I can attempt this. Although, last time I tried this (months, maybe up to a year ago) it rendered some of the custom nodes I depended on, unable to function. It will take some time to uninstall PyTorch, install PyTorch, and test all of my workflows, but I'll give it a go. I have a few questions.

As a question for contingency, do you think the the Legacy loader can be patched? Or do you see enabling/backporting Dynamic VRAM being the only possibility? I do need to keep v0.28.0 for the time being, so I'm assuming I'd also need to '--enable-dynamic-vram' as well if v0.28.0 predates Dynamic VRAM functionality?

molbal

Owner 5 days ago

I'll look into it if it's possible, I'm not sure without checking the code, and I'm not in front of my PC now. Alternatively, you can see if you can download and load a regular INT8 Safetensors checkpoint, that should have equivalent performance to the Q8_CR GGUF and might work without changing your setup

psdboi

5 days ago

Thanks, that's all I can ask for, if the possibility is there. I've tried INT8/ConvRot models in the past, and for whatever reason they perform worse than GGUF (I guess it's just me/my setup). That's until I discovered your models because the Krea2 Turbo Q8_CR, aside from the LoRA issue, worked perfectly as I would have imagined; reduced VRAM and a slight bump in generation speed. So I'm a big fan of the .gguf format which is why I'm so determined to get this CR variant to work. But, I'm still working on the PyTorch swap. It's just a pain; change one installation and have to change a few others for everything to work again.

psdboi

5 days ago

•

edited 5 days ago

I've updated to PyTorch: 2.13.0+cu130

Here are my results:

Dynamic VRAM Node: Krea2 Turbo Q8_CR with LoRA *OFF:

Dynamic VRAM Node: Krea2 Turbo Q8_CR with LoRA *ON:

Dynamic VRAM Node: Krea2 Turbo Q8_CR with LoRA *ON (4-Steps):

Legacy Node: Krea2 Turbo Q8_CR with LoRA *OFF:

Legacy Node: Krea2 Turbo Q8_CR with LoRA *ON:

Dynamic VRAM Node: Krea2 Turbo Q8_0 with LoRA *OFF:

Dynamic VRAM Node: Krea2 Turbo Q8_0 with LoRA *ON:

Dynamic VRAM Node: Krea2 Turbo Q8_0 with LoRA *ON (4-Steps):

Legacy Node: Krea2 Turbo Q8_0 with LoRA *OFF:

Legacy Node: Krea2 Turbo Q8_0 with LoRA *ON:

What I can say is: Q8_0 is still the only one that applies the LoRA correctly (Legacy node only). Now Q8_CR runs in Dynamic VRAM, but it produces a half-cooked/noisy image regardless of the GGUF node. Interestingly enough, I tried running Q8_CR and Q8_0 with 16 and 4 steps. And 16 was just a blob and 4, surprisingly, got close to the target but not exact; less than accurate style quality.

Also, as I suspected. The PyTorch upgrade broke a number of my custom nodes compatibility. Unfortunately, it's more than I can deal with today; too many errors to source through and too many fixes required to run the workflows. For now, it seems I'll just have to use Q8_0 in the Legacy Node. Maybe revert PyTorch changes if it's too bothersome, as well. At least until something else pans out.

molbal

Owner 4 days ago

Thanks this is very detailed. I'll take a look if I can add support for the legacy loader node.

psdboi

4 days ago

Awesome, thanks for considering it.

molbal

Owner 4 days ago

Would update the nodes at least to https://github.com/molbal/ComfyUI-GGUF/commit/967b26c777c32ad61ccc4b76c97cd6c1d1e3e203 (or just the latest in the main branch) and test again please?

I added a fix and now LoRAs apply also on the legacy vram management path, however it is considerably slower on my machine.

psdboi

3 days ago

•

edited 3 days ago

Okay, I'll look a it and let you know what I find.

Edit: (Question) Which version of ComfyUI are you using? I want to try that as well to make sure we're synced on versions. I can perform a git checkout, swap, and do A/B test of 0.28.0 and your version.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment