Access Violation (0xC0000005) in convert_diffusers_mmdit when loading FLUX Fill model after v0.15.1

[rong7932]

Custom Node Testing

• I have tried disabling custom nodes and the issue persists (see how to disable custom nodes if you need help)

Expected Behavior

The FLUX.1-Fill-Dev model (specifically checkpoints like flux.1-fill-dev-OneReward) should load successfully and enter the sampling stage. It worked perfectly in v0.15.1 (Commit: 3dd10a5).

Actual Behavior

The process crashes immediately with a Windows Access Violation (0xC0000005) during the model weight conversion stage (convert_diffusers_mmdit). In some cases, it is preceded by an aiohttp TypeError. The backend process exits completely (Code: 3221225477).

Steps to Reproduce

Update ComfyUI to any version newer than v0.15.1 (Commit: 3dd10a5).

Use Python 3.12 or 3.13 on a Windows environment.

Use a basic workflow to load the flux.1-fill-dev-OneReward checkpoint (or any standard FLUX Fill model).

Queue the prompt.

The crash occurs exactly when the console reaches the MMDIT weight conversion stage.

Note: Standard FLUX.1 Dev/Schnell models load correctly; only the Fill architecture triggers this crash.

Debug Logs

FETCH ComfyRegistry Data: 125/129
FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
 [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
fast DB scan failed for models: 'NoneType' object is not callable
Traceback (most recent call last):
  File "D:\AI\ComfyUINeo\app\assets\scanner.py", line 36, in seed_assets
    survivors: set[str] = _fast_db_consistency_pass(r, collect_existing_paths=True, update_missing_tags=True)
                          ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\AI\ComfyUINeo\app\assets\scanner.py", line 156, in _fast_db_consistency_pass
    with create_session() as sess:
         ~~~~~~~~~~~~~~^^
  File "D:\AI\ComfyUINeo\app\database\db.py", line 112, in create_session
    return Session()
TypeError: 'NoneType' object is not callable
orphan pruning failed: 'NoneType' object is not callable
Traceback (most recent call last):
  File "D:\AI\ComfyUINeo\app\assets\scanner.py", line 43, in seed_assets
    orphans_pruned = _prune_orphaned_assets(roots)
  File "D:\AI\ComfyUINeo\app\assets\scanner.py", line 123, in _prune_orphaned_assets
    with create_session() as sess:
         ~~~~~~~~~~~~~~^^
  File "D:\AI\ComfyUINeo\app\database\db.py", line 112, in create_session
    return Session()
TypeError: 'NoneType' object is not callable
Failed to seed assets: 'NoneType' object is not callable
got prompt
Using xformers attention in VAE
Using xformers attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
Found quantization metadata version 1
Using MixedPrecisionOps for text encoder
clip missing: ['text_projection.weight']
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load FluxClipModel_
Model FluxClipModel_ prepared for dynamic VRAM loading. 4903MB Staged. 0 patches attached. Force pre-loaded 49 weights: 392 KB.
Requested to load AutoencodingEngine
Model AutoencodingEngine prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model AutoencodingEngine prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Windows fatal exception: access violation

Windows fatal exception: Thread 0x00008f8c (most recent call first):
  File "Windows fatal exception: D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in access violation_workerWindows fatal exception: 
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line Windows fatal exception: access violation994access violation

 in 

run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00006a78 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run


  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00008bfc (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py
", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00002fcc (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py"Windows fatal exception: , line 90 in _worker
Windows fatal exception: access violation

Windows fatal exception: access violation

Windows fatal exception: access violation

access violation

access violation

Windows fatal exception: access violation

  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00009c18 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py"
, line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00009344 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00002cf4 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00009f80 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\concurrent\futures\thread.py", line 90 in _worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x000093f0 (most recent call first):
  File "D:\AI\ComfyUINeo\comfy\model_detection.py", line 1125 in convert_diffusers_mmdit
  File "D:\AI\ComfyUINeo\comfy\sd.py", line 1717 in load_diffusion_model_state_dict
  File "D:\AI\ComfyUINeo\comfy\sd.py", line 1771 in load_diffusion_model
  File "D:\AI\ComfyUINeo\nodes.py", line 972 in load_unet
  File "D:\AI\ComfyUINeo\execution.py", line 295 in process_inputs
  File "D:\AI\ComfyUINeo\execution.py", line 307 in _async_map_node_over_list
  File "D:\AI\ComfyUINeo\custom_nodes\ComfyUI-Lora-Manager\py\metadata_collector\metadata_hook.py", line 168 in async_map_node_over_list_with_metadata
  File "D:\AI\ComfyUINeo\execution.py", line 333 in get_output_data
  File "D:\AI\ComfyUINeo\execution.py", line 524 in execute
  File "D:\AI\ComfyUINeo\custom_nodes\ComfyUI-Lora-Manager\py\metadata_collector\metadata_hook.py", line 203 in async_execute_with_prompt_tracking
  File "D:\AI\ComfyUINeo\execution.py", line 737 in execute_async
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\events.py", line 89 in _run
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 2050 in _run_once
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 683 in run_forever
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 712 in run_until_complete
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\runners.py", line 118 in run
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\runners.py", line 195 in run
  File "D:\AI\ComfyUINeo\execution.py", line 688 in execute
  File "D:\AI\ComfyUINeo\main.py", line 261 in prompt_worker
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00006628 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 359 in wait
  File "D:\AI\ComfyUINeo\py313\Lib\queue.py", line 199 in get
  File "D:\AI\ComfyUINeo\custom_nodes\ComfyUI_VNCCS_Utils\nodes\vnccs_model_manager.py", line 116 in worker_loop
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00001278 (most recent call first):
  File "D:\AI\ComfyUINeo\custom_nodes\comfyui_fk_server\server.py", line 59 in _run_node_program
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00001c18 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\windows_events.py", line 775 in _poll
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\windows_events.py", line 446 in select
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 2012 in _run_once
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 683 in run_forever
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 712 in run_until_complete
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\runners.py", line 118 in run
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\runners.py", line 195 in run
  File "D:\AI\ComfyUINeo\custom_nodes\ComfyUI-Crystools\general\monitor.py", line 31 in startMonitorLoop
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 994 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00009430 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 359 in wait
  File "D:\AI\ComfyUINeo\py313\Lib\queue.py", line 199 in get
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\api.py", line 379 in dispatch_events
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\api.py", line 213 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x00007bb0 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\winapi.py", line 326 in read_directory_changes
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\winapi.py", line 380 in read_events
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\read_directory_changes.py", line 67 in _read_events
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\read_directory_changes.py", line 70 in queue_events
  File "D:\AI\ComfyUINeo\py313\Lib\site-packages\watchdog\observers\api.py", line 158 in run
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1043 in _bootstrap_inner
  File "D:\AI\ComfyUINeo\py313\Lib\threading.py", line 1014 in _bootstrap

Thread 0x000078b8 (most recent call first):
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\windows_events.py", line 775 in _poll
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\windows_events.py", line 446 in select
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 2012 in _run_once
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 683 in run_forever
  File "D:\AI\ComfyUINeo\py313\Lib\asyncio\base_events.py", line 712 in run_until_complete
  File "D:\AI\ComfyUINeo\main.py", line 440 in <module>

Other

Last Working Commit: 3dd10a5 (v0.15.1, Feb 26, 2026).

Environment: Windows 11, RTX 5060 Ti (16GB), 64GB RAM.

Hypothesis: The regression seems to be in the new tensor offset calculation logic for models with extra input channels (like FLUX Fill) within the convert_diffusers_mmdit function.

Even with --disable-smart-memory and --deterministic, the issue persists on the latest master.

[gangdeq]

我也出现了同样的情况，FLUX Fill OneReward这里fill模型会发生崩溃， #12854

[rong7932]

我也出现了同样的情况，FLUX Fill OneReward这里fill模型会发生崩溃，#12854

这个是内核更新之后出现的bug。所有使用这个模型的都会。除非降级到0.15.1版本。这是我试过最后可以用的版本。

[CHKE9]

am i in same boat? python crashing

pc specs
RTX 5090
64gb system ram

Model info:
While running an infinite talk lasting about 40 seconds of video, this error occurred the second time the workflow was executed, but not during the first run.

Event Viewer log:
Faulting application name: python.exe, version: 3.12.10150.1013, time stamp: 0x67f515a7
Faulting module name: c10.dll, version: 0.0.0.0, time stamp: 0x690e8612
Exception code: 0xc0000005
Fault offset: 0x000000000008bf55
Faulting process id: 0x944C
Faulting application start time: 0x1DCB0307E246A81
Faulting application path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\python.exe
Faulting module path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\lib\c10.dll
Report Id: b38a29b9-11af-47fe-8984-fcf5ed8ca08c
Faulting package full name:
Faulting package-relative application ID:

[rattus128]

am i in same boat? python crashing

pc specs RTX 5090 64gb system ram

Model info: While running an infinite talk lasting about 40 seconds of video, this error occurred the second time the workflow was executed, but not during the first run.

Event Viewer log: Faulting application name: python.exe, version: 3.12.10150.1013, time stamp: 0x67f515a7 Faulting module name: c10.dll, version: 0.0.0.0, time stamp: 0x690e8612 Exception code: 0xc0000005 Fault offset: 0x000000000008bf55 Faulting process id: 0x944C Faulting application start time: 0x1DCB0307E246A81 Faulting application path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\python.exe Faulting module path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\lib\c10.dll Report Id: b38a29b9-11af-47fe-8984-fcf5ed8ca08c Faulting package full name: Faulting package-relative application ID:

No that looks different. Can I get the workflow and console logs leading up to the crash? Are you able to reproduce with something minimal as well and does it reproduce without custom nodes?

[jprsyt5]

Model Lumina2 prepared for dynamic VRAM loading. 11739MB Staged. 0 patches attached.
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:05<00:00,  1.29s/it]
Requested to load Lumina2
Model Lumina2 prepared for dynamic VRAM loading. 11739MB Staged. 180 patches attached.
 12%|██████▏                                          | 1/8 [00:03<00:23,  3.31s/it,  Model Initialization complete!  ]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 154
 25%|█████████████████████                                                               | 2/8 [00:04<00:14,  2.39s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 152
 38%|███████████████████████████████▌                                                    | 3/8 [00:07<00:11,  2.31s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 152
 50%|██████████████████████████████████████████                                          | 4/8 [00:09<00:09,  2.37s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 148
 62%|████████████████████████████████████████████████████▌                               | 5/8 [00:11<00:07,  2.40s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 144
 75%|███████████████████████████████████████████████████████████████                     | 6/8 [00:14<00:04,  2.43s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143
 88%|█████████████████████████████████████████████████████████████████████████▌          | 7/8 [00:16<00:02,  2.39s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143
100%|████████████████████████████████████████████████████████████████████████████████████| 8/8 [00:19<00:00,  2.39s/it]
0 models unloaded.
Model AutoencodingEngine prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143

@rattus128 sorry for jumping in here. This is my first time seeing this after a few weeks of keeping up with the latest ComfyUI & aimdo updates, and I don’t think this is really an issue (?) so I’m not sure if I should open a separate issue ticket, even when the warning shows up, everything still works fine. I’m just wondering what this warning is about?

And I’m not sure how to reproduce it, but it sometimes shows up after a few runs. And then, on the next run it might not appear again.

The aimdo logs definitely don’t look like human-readable language lol

[paul-lupu]

Normal fill seems to still work for me, but flux srpo and fill-one-reward crash comfy as mentioned here: #12873

[rong7932]

普通填充对我来说似乎还能用，但flux srpo和填充一次奖励的崩溃效果很舒服，正如这里提到的：#12873

I mean that the flux-fill-onereward model will crash when it loads, and other models will not

[CHKE9]

am i in same boat? python crashing
pc specs RTX 5090 64gb system ram
Model info: While running an infinite talk lasting about 40 seconds of video, this error occurred the second time the workflow was executed, but not during the first run.
Event Viewer log: Faulting application name: python.exe, version: 3.12.10150.1013, time stamp: 0x67f515a7 Faulting module name: c10.dll, version: 0.0.0.0, time stamp: 0x690e8612 Exception code: 0xc0000005 Fault offset: 0x000000000008bf55 Faulting process id: 0x944C Faulting application start time: 0x1DCB0307E246A81 Faulting application path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\python.exe Faulting module path: D:\Comfy\Comfy-Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\lib\c10.dll Report Id: b38a29b9-11af-47fe-8984-fcf5ed8ca08c Faulting package full name: Faulting package-relative application ID:

No that looks different. Can I get the workflow and console logs leading up to the crash? Are you able to reproduce with something minimal as well and does it reproduce without custom nodes?

unable to reproduce it its random, when you let it run two-three times it it will crash, is it due to pagefile?
BTW Flux Fill OneReward works fine for me

Wan InfiniteTalk- Easy Extend from 3 to 34 Seconds.json

[rattus128]

@CKay010 we just merged a PR that will improve the logging on these crashes so if you update to the latest git master, if it crashes we should get a much more informative log.

[rattus128]

For anyone tracking Flux Fill OneReward, this is fixed in the latest git master. Please pull and retest @paul-lupu @rong7932 or retest the next stable release. --disable-dynamic-vram will workaround for the next day or so if you just want to wait.

[rattus128]

Model Lumina2 prepared for dynamic VRAM loading. 11739MB Staged. 0 patches attached.
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:05<00:00, 1.29s/it]
Requested to load Lumina2
Model Lumina2 prepared for dynamic VRAM loading. 11739MB Staged. 180 patches attached.
12%|██████▏ | 1/8 [00:03<00:23, 3.31s/it, Model Initialization complete! ]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 154
25%|█████████████████████ | 2/8 [00:04<00:14, 2.39s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 152
38%|███████████████████████████████▌ | 3/8 [00:07<00:11, 2.31s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 152
50%|██████████████████████████████████████████ | 4/8 [00:09<00:09, 2.37s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 148
62%|████████████████████████████████████████████████████▌ | 5/8 [00:11<00:07, 2.40s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 144
75%|███████████████████████████████████████████████████████████████ | 6/8 [00:14<00:04, 2.43s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143
88%|█████████████████████████████████████████████████████████████████████████▌ | 7/8 [00:16<00:02, 2.39s/it]aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143
100%|████████████████████████████████████████████████████████████████████████████████████| 8/8 [00:19<00:00, 2.39s/it]
0 models unloaded.
Model AutoencodingEngine prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
aimdo: src/model-vbar.c:65:WARNING:VBAR 000002DC18071F90: Resident page 166 is ABOVE watermark 143

@rattus128 sorry for jumping in here. This is my first time seeing this after a few weeks of keeping up with the latest ComfyUI & aimdo updates, and I don’t think this is really an issue (?) so I’m not sure if I should open a separate issue ticket, even when the warning shows up, everything still works fine. I’m just wondering what this warning is about?

And I’m not sure how to reproduce it, but it sometimes shows up after a few runs. And then, on the next run it might not appear again.

The aimdo logs definitely don’t look like human-readable language lol

This usually isnt a functional issue. It indicates a timing interaction where some cleanup has to be delayed due to sections of VRAM getting frozen for use vs the allocator trying to make space and a critical section of VRAM remains stranded as an allocation while being marked for free.

Thanks for the report. Ill see if I can find a root cause.