commit 916fc0b1bcfd37b6bd9ece0daeb5b3cbaa53d0a9
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Mon Dec 15 17:30:37 2025 +0200
Update nodes.py
commit 63818324f5dbb0b300064bea0402c4cd1bd57b2b
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Mon Dec 15 17:30:26 2025 +0200
Refactor RoPE caching
commit bb0c55da4d8f8bca4968704e877fd057a90a1eeb
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Mon Dec 15 01:59:16 2025 +0200
Update nodes_sampler.py
commit a0447d55534857051606ee4201bc7f4e25aa73ae
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Mon Dec 15 01:28:09 2025 +0200
Fix non scale wfs
commit fa761cc2f2a426faa9c391aeede62cf6f0fd7266
Merge: ea1677b 3aae54f
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Mon Dec 15 01:26:23 2025 +0200
Merge branch 'main' into SCAIL
commit ea1677bd4ad42f19e369551590a9d4f17a36fa29
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 19:41:43 2025 +0200
Handle torchscript issue better
Some other custom nodes globally set torch._C._jit_set_profiling_executor(False) which breaks the NLF model
commit e3cfa64bd3712ac153ce84a75215842c884a8ba4
Merge: ad7a0b9 3611341
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 16:49:04 2025 +0200
Merge branch 'main' into SCAIL
commit ad7a0b925de61ff705b928cd802e752e46089b42
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 16:10:34 2025 +0200
Fix possible uni3c issue
commit 74d97fa4bb7c58a0edf8516cc9fad4468da5c57e
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 15:58:42 2025 +0200
Match Uni3C temporal dim
commit 056d8ad96ffa5a223a8cd88c900a573a8d450e22
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 14:47:58 2025 +0200
Add warning for potential other overrides on torch.jit.script
commit f6dff002ffdcd880451955db298872ea90a4e3f8
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 14:19:33 2025 +0200
Add option to warmup the NLF model on load and fix it's offloading
commit a19107501dff23804e7db984d7da304a9955adc9
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sun Dec 14 13:45:20 2025 +0200
Add error to indicate ComfyUI-RMBG currently breaks the NLF model
commit e2cfa486e48ead50195884167d9794c7caf0a69f
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 23:29:49 2025 +0200
Cleanup unnecessary code
commit 462b61855fb96b0cb18cbccd48593256992808d7
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 18:05:10 2025 +0200
context windows
commit e57d4baeebf12c43e851c6c2467d698d7dbb4d03
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 16:55:23 2025 +0200
Start/end percentages and strength
commit 3e507ae32256ed3e41cea69d9e26c30b5272968e
Merge: 1e5c7cb 0fa5383
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 16:09:16 2025 +0200
Merge branch 'main' into SCAIL
commit 1e5c7cb2113138bdeae562d266f911c1e3edee91
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 15:45:39 2025 +0200
Update nodes.py
commit 98f8e56bcacfc07e12cbb4b26555b2b28d9db92f
Merge: 9652146 78e3e18
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 15:42:44 2025 +0200
Merge branch 'main' into SCAIL
commit 9652146763fb27e916a6853a8125efd0a67cd601
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 02:41:06 2025 +0200
Add imitation of SCAIL pose drawing to the existing NLF node
This only draws the pose with same colors, it's not meant as final solution, just for testing.
commit 1f86cebdaa97570ed88da0c9986b85c6664d62dc
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Sat Dec 13 01:11:56 2025 +0200
test pose inputs
commit b348b21dbef0dcb92c0961df85959648e78da6aa
Author: kijai <40791699+kijai@users.noreply.github.com>
Date: Fri Dec 12 20:10:48 2025 +0200
Init
Fixed RuntimeError when using I2V embeds with Lynx embeds where tensors
were on different devices (cuda:0 and cpu).
The issue occurred because lynx_ref_text_embed["prompt_embeds"] were not
explicitly moved to the GPU device before being passed to the transformer
during Lynx reference buffer extraction.
Changes:
- Move lynx text embeddings to device in both conditional and unconditional
buffer extraction calls (lines 1114 and 1128)
- Ensures all tensors are on the same device during cross-attention operations
This change has caused issues especially with LoRAs that have dynamic rank. Will now be disabled by default, to allow full graph with unmerged LoRAs the option to allow compile is available in the Torch Compile Settings -node
kijai