3fcdac8c7b
bunch of fixes
2026-02-22 23:47:49 +02:00
49d1eab2a5
fixes
2026-02-22 01:25:10 +02:00
c000fe090f
debugging
2026-02-20 22:04:37 +02:00
167ab754d9
coderabbit 2
2026-02-20 21:13:13 +02:00
acb25b55e8
code rabbit suggestions
2026-02-20 20:16:49 +02:00
8e1230543d
.
2026-02-20 17:39:44 +02:00
578c64f2c4
trellis2conditioning and a hidden bug
2026-02-19 22:05:33 +02:00
4e401fd353
small fixes
2026-02-19 00:48:26 +02:00
f2d477720c
.
2026-02-18 22:01:09 +02:00
5a51bf82de
Merge branch 'master' into trellis2
2026-02-18 21:52:39 +02:00
44f8598521
Fix anima LLM adapter forward when manual cast ( #12504 )
2026-02-17 07:56:44 -08:00
18927538a1
Implement NAG on all the models based on the Flux code. ( #12500 )
...
Use the Normalized Attention Guidance node.
Flux, Flux2, Klein, Chroma, Chroma radiance, Hunyuan Video, etc..
2026-02-16 23:30:34 -05:00
b7de0ad87e
postprocessing node fixes + model small fixes
2026-02-17 00:10:48 +02:00
88e6370527
Remove workaround for old pytorch. ( #12480 )
2026-02-15 20:43:53 -05:00
ee24b93abc
rewriting conditioning logic + model code addition
2026-02-16 01:53:53 +02:00
dc9822b7df
Add working Qwen 2512 ControlNet (Fun ControlNet) support ( #12359 )
2026-02-13 22:23:52 -05:00
726af73867
Fix some custom nodes. ( #12455 )
2026-02-13 20:21:10 -05:00
e1add563f9
Use torch RMSNorm for flux models and refactor hunyuan video code. ( #12432 )
2026-02-13 15:35:13 -05:00
6023feca0e
.
2026-02-13 21:05:59 +02:00
3309aaafd0
more reliable detection
2026-02-13 00:10:25 +02:00
9b9c3c3b9f
debugging
2026-02-11 20:33:59 +02:00
76a7fa96db
Make built in lora training work on anima. ( #12402 )
2026-02-10 22:04:32 -05:00
cdcf4119b3
[Trainer] training with proper offloading ( #12189 )
...
* Fix bypass dtype/device moving
* Force offloading mode for training
* training context var
* offloading implementation in training node
* fix wrong input type
* Support bypass load lora model, correct adapter/offloading handling
2026-02-10 21:45:19 -05:00
f9709a4ee6
dinov3 fixes + other
2026-02-11 01:27:54 +02:00
b0b2273b4f
small bug fixes
2026-02-09 00:41:01 +02:00
039955c527
Some fixes to previous pr. ( #12339 )
2026-02-06 20:14:52 -05:00
6a26328842
Support fp16 for Cosmos-Predict2 and Anima ( #12249 )
2026-02-06 20:12:15 -05:00
eba6c940fd
Make ace step 1.5 base model work properly with default workflow. ( #12337 )
2026-02-06 19:14:56 -05:00
1b73bd4b50
post-process node
2026-02-06 23:54:27 +02:00
d6006f62dd
checkpoint
2026-02-06 20:35:33 +02:00
bc02ecfb9f
model fixes
2026-02-06 19:28:49 +02:00
458292fef0
Fix some lowvram stuff with ace step 1.5 ( #12312 )
2026-02-05 19:15:04 -05:00
6555dc65b8
Make ace step 1.5 work without the llm. ( #12311 )
2026-02-05 16:43:45 -05:00
57830e08f0
fixes to vae and cumesh impl.
2026-02-05 17:19:57 +02:00
a50c32d63f
Disable sage attention on ace step 1.5 ( #12297 )
2026-02-04 22:15:30 -05:00
6125b80979
Add llm sampling options and make reference audio work on ace step 1.5 ( #12295 )
2026-02-04 21:29:22 -05:00
0ed8f5c925
apply rope and optimized attention
2026-02-05 02:34:08 +02:00
2865e385e2
needed updates
2026-02-04 14:15:00 +02:00
8d9faf28e6
structure model
2026-02-03 22:40:54 +02:00
2b5b54ae83
model init working
2026-02-03 21:10:20 +02:00
3c1a1a2df8
Basic support for the ace step 1.5 model. ( #12237 )
2026-02-03 00:06:18 -05:00
9a2c01d160
..
2026-02-02 21:27:15 +02:00
da84a2b868
Merge branch 'master' into trellis2
2026-02-02 18:13:58 +02:00
f8acd9c402
Reduce RAM usage, fix VRAM OOMs, and fix Windows shared memory spilling with adaptive model loading ( #11845 )
2026-02-01 01:01:11 -05:00
b8f848bfe3
Fix model not working with any res. ( #12186 )
2026-01-31 00:12:48 -05:00
6ea2e5b288
init
2026-01-30 23:34:48 +02:00
6516ab335d
wan-vae: Switch off feature cache for single frame ( #12090 )
...
The code throughout is None safe to just skip the feature cache saving
step if none. Set it none in single frame use so qwen doesn't burn VRAM
on the unused cache.
2026-01-26 19:40:19 -05:00
635406e283
Only enable fp16 on z image models that actually support it. ( #12065 )
2026-01-24 22:32:28 -05:00
4e6a1b66a9
speed up and reduce VRAM of QWEN VAE and WAN (less so) ( #12036 )
...
* ops: introduce autopad for conv3d
This works around pytorch missing ability to causal pad as part of the
kernel and avoids massive weight duplications for padding.
* wan-vae: rework causal padding
This currently uses F.pad which takes a full deep copy and is liable to
be the VRAM peak. Instead, kick spatial padding back to the op and
consolidate the temporal padding with the cat for the cache.
* wan-vae: implement zero pad fast path
The WAN VAE is also QWEN where it is used single-image. These
convolutions are however zero padded 3d convolutions, which means the
VAE is actually just 2D down the last element of the conv weight in
the temporal dimension. Fast path this, to avoid adding zeros that
then just evaporate in convoluton math but cost computation.
2026-01-23 19:56:14 -05:00
55bd606e92
LTX2: Refactor forward function for better VRAM efficiency and fix spatial inpainting ( #12046 )
...
* Disable timestep embed compression when inpainting
Spatial inpainting not compatible with the compression
* Reduce crossattn peak VRAM
* LTX2: Refactor forward function for better VRAM efficiency
2026-01-23 15:26:38 -05:00
Yousef Rafat
chaObserv
comfyanonymous
krigeta
Kohaku-Blueleaf
tdrussell
rattus
Yousef R. Gamaleldin
Jukka Seppänen