search

comfyanonymous / ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
https://www.comfy.org/
GNU General Public License v3.0
50.63k stars 5.32k forks source link

Workflow runs once since updating 6/26/2024 #3890

Closed pryingopen3rdi closed 2 months ago

pryingopen3rdi commented 2 months ago

Expected Behavior

I have been running this same workflow for awhile now. I expect to be able to run more than one batch.

Actual Behavior

When I start up Comfyui and run my workflow it runs fine for the first run. After that it only runs the Vae Decode or it won't let me queue up at all.

Steps to Reproduce

Load workflow , run to stage 3. Try to queue another time. workflow_demoncore4.1_V1.0.json

Debug Logs

[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2024-06-27 06:28:42.846144
** Platform: Linux
** Python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]
** Python executable: /usr/bin/python3
** Log path: /home/zaran/AIDrive/ComfyUI/comfyui.log

Prestartup times for custom nodes:
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/rgthree-comfy
   0.6 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Manager

Total VRAM 24116 MB, total RAM 64208 MB
pytorch version: 2.4.0.dev20240609+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync
Using pytorch cross attention
/usr/lib/python3/dist-packages/requests/__init__.py:87: RequestsDependencyWarning: urllib3 (2.2.2) or chardet (4.0.0) doesn't match a supported version!
  warnings.warn("urllib3 ({}) or chardet ({}) doesn't match a supported "
/home/zaran/.local/lib/python3.10/site-packages/kornia/feature/lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
  @torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
Adding /home/zaran/AIDrive/ComfyUI/custom_nodes to sys.path
Loaded Efficiency nodes from /home/zaran/AIDrive/ComfyUI/custom_nodes/efficiency-nodes-comfyui
Could not find ControlNetPreprocessors nodes
Loaded AdvancedControlNet nodes from /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Advanced-ControlNet
Could not find AnimateDiff nodes
Loaded IPAdapter nodes from /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_IPAdapter_plus
Could not find VideoHelperSuite nodes
### Loading: ComfyUI-Impact-Pack (V5.18.3)
### Loading: ComfyUI-Impact-Pack (Subpack: V0.6)
### Loading: ComfyUI-Impact-Pack (V5.18.3)
Loaded ImpactPack nodes from /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Impact-Pack
[Impact Pack] Wildcards loading done.
[Impact Pack] Wildcards loading done.
### Loading: ComfyUI-Manager (V2.42)
### ComfyUI Revision: 2295 [4f9d2b05] | Released on '2024-06-27'
### Loading: ComfyUI-Inspire-Pack (V0.80.4)
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
WAS Node Suite: BlenderNeko's Advanced CLIP Text Encode found, attempting to enable `CLIPTextEncode` support.
WAS Node Suite: `CLIPTextEncode (BlenderNeko Advanced + NSP)` node enabled under `WAS Suite/Conditioning` menu.
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
WAS Node Suite: OpenCV Python installed.
WAS Node Suite: Finished. Loaded 217 nodes successfully.

        "Art is the most beautiful of all lies." - Claude Debussy

### Loading: ComfyUI-Impact-Pack (V5.18.3)
[Impact Pack] Wildcards loading done.
Total VRAM 24116 MB, total RAM 64208 MB
pytorch version: 2.4.0.dev20240609+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync

[rgthree] Loaded 40 fantastic nodes.
[rgthree] Will use rgthree's optimized recursive execution.

------------------------------------------
Comfyroll Studio v1.76 :  175 Nodes Loaded
------------------------------------------
** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md
** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki
------------------------------------------

Import times for custom nodes:
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Advanced-ControlNet
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/websocket_image_save.py
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/efficiency-nodes-comfyui
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/LoadLoraWithTags
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/comfyui-upscale-by-model
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/sd-perturbed-attention
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_ADV_CLIP_emb
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/sd-dynamic-thresholding
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/lora-info
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_restart_sampling
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-post-processing-nodes
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Impact-Pack
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_UltimateSDUpscale
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/rgthree-comfy
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-KJNodes
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_IPAdapter_plus
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI_Comfyroll_CustomNodes
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Manager
   0.0 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-Inspire-Pack
   0.5 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/comfyui-art-venture
   1.1 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/was-node-suite-comfyui
   4.4 seconds: /home/zaran/AIDrive/ComfyUI/custom_nodes/ComfyUI-StableAudioSampler

Starting server

To see the GUI go to: http://127.0.0.1:8188
got prompt
[rgthree] Using rgthree's optimized recursive execution.
[rgthree] First run patching recursive_output_delete_if_changed and recursive_will_execute.
[rgthree] Note: If execution seems broken due to forward ComfyUI changes, you can disable the optimization from rgthree settings in ComfyUI.
model_type EPS
Using pytorch attention in VAE
Using pytorch attention in VAE
Requested to load SDXLClipModel
Loading 1 new model
Requested to load SDXL
Loading 1 new model
100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 25/25 [00:06<00:00,  3.69it/s]
Using pytorch attention in VAE
Using pytorch attention in VAE
Requested to load AutoencoderKL
Loading 1 new model
lora key not loaded: lora_te1_text_model_encoder_layers_0_mlp_fc1.alpha
//NOTE The above part happens over and over, removed to save space//
[Info] CR Apply Multi Upscale: Applying 4x_foolhardy_Remacri.pth and rescaling by factor 1.5
WARNING: comfy_extras.chainner_models is deprecated and has been replaced by the spandrel library.
Requested to load SDXL
Loading 1 new model
100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 15/15 [00:08<00:00,  1.69it/s]
WARNING: comfy_extras.chainner_models is deprecated and has been replaced by the spandrel library.
Prompt executed in 45.60 seconds
got prompt
[rgthree] Using rgthree's optimized recursive execution.
Prompt executed in 0.00 seconds
got prompt
[rgthree] Using rgthree's optimized recursive execution.
Prompt executed in 0.00 seconds

Other

No response

pryingopen3rdi commented 2 months ago

Another note. It will sometimes run more than once, if I'm lucky but usually dies after the first attempt.

pryingopen3rdi commented 2 months ago

When it actually fails there is no error message. The prompt just finishes quicker because it only runs the VaeDecodeTiled process and nothing else.

got prompt [rgthree] Using rgthree's optimized recursive execution. Prompt executed in 4.21 seconds

pryingopen3rdi commented 2 months ago

Reinstalled comfy in new VENV and only the nodes required for workflow and still having the same issue.

mcmonkey4eva commented 2 months ago

that workflow is too complex and full of custom nodes to check, but -- check if you have all your seeds set to "fixed", any fixed seed inputs intentionally don't re-run (since the output would be identical)

pryingopen3rdi commented 2 months ago

So, none of them were fixed but a few were set to incremental. I changed that and the issue resolved. Still strange that it didn't start happening until a few days ago. Thank you, you can close this.

writingdeveloper commented 1 month ago

I also have same problem, and I realized that everytime we runs, seeds should be change. If not comfyUI will pass the execute process.

I add this options in 'inputs' object

"control_after_generate":"randomize"

ltdrdata commented 1 month ago

I also have same problem, and I realized that everytime we runs, seeds should be change. If not comfyUI will pass the execute process.

I add this options in 'inputs' object

"control_after_generate":"randomize"

ComfyUI's default behavior is to use cache if the prompt doesn't change at all. If settings of some nodes in the workflow change, only those nodes and the subsequent affected nodes are re-executed. This is the fundamental principle of the workflow approach. If every workflow were to always re-execute without any changes, it would be a terrible situation when dealing with large workflows.