kijai
3cefbf0d70
small updates
2025-03-06 02:01:24 +02:00
kijai
f4e706156f
Update nodes.py
2025-03-05 22:07:36 +02:00
kijai
32465c2787
TeaCache tweaks for context windowing and FlowEdit
2025-03-05 21:02:34 +02:00
kijai
de3e0f9ef2
update TeaCache
2025-03-05 13:35:18 +02:00
kijai
43dc9a76c1
Update nodes.py
2025-03-05 08:55:42 +02:00
kijai
c4472c2906
Add FlowEdit and some code refactors
2025-03-05 00:54:08 +02:00
kijai
721cd65e7b
Apply v2v latents after freenoise
2025-03-04 13:41:29 +02:00
kijai
5c037973aa
update RifleX default freq
2025-03-04 11:27:10 +02:00
kijai
53710b0fea
Update nodes.py
2025-03-04 11:15:16 +02:00
kijai
013fae65fb
cleanup
2025-03-04 01:56:13 +02:00
kijai
8c451f4a2f
fix context windows
2025-03-03 01:40:33 +02:00
kijai
0183a03bf9
revert and fix again...
...
I'll clean this later
2025-03-03 01:38:51 +02:00
kijai
24fb1b42fe
bug fixes
2025-03-03 01:35:36 +02:00
kijai
8e0a90f1a4
handle I2V a bit for contexting
2025-03-02 20:09:23 +02:00
kijai
bd31044ef5
Experimental: Allow multiple prompts for context windows
2025-03-02 19:48:54 +02:00
kijai
7697159991
Update nodes.py
2025-03-02 17:39:30 +02:00
kijai
f41f10a4ee
fix text embed casting
2025-03-02 17:16:29 +02:00
kijai
0479032aac
Update nodes.py
2025-03-02 17:03:54 +02:00
kijai
bf2511fbb0
Allow using ComfyUI native clip_vision
2025-03-02 16:57:34 +02:00
kijai
5b9ee2bc7e
Add a way to use native comfy text encoding
2025-03-02 16:35:54 +02:00
kijai
2dba474882
Update nodes.py
2025-03-02 15:27:23 +02:00
kijai
d3129c6996
Add alternative offloading method from Diffsynth
2025-03-02 14:37:08 +02:00
kijai
55dcc33d00
Add block swap debug prints and cleanup unused code
2025-03-02 10:49:46 +02:00
kijai
9087791b75
Allow setting TeaCache device to avoid the VRAM usage increase
2025-03-02 09:57:02 +02:00
kijai
b30f4584b7
fix previous fix
2025-03-02 02:22:49 +02:00
kijai
df59175ac4
fix feta for context windows
2025-03-02 02:14:07 +02:00
kijai
fa3d61cb45
Cleanup the text part of the clip model as redundant as it's not even used
2025-03-02 01:44:19 +02:00
kijai
7593e26457
fix TeaCache for context windows (maybe)
2025-03-02 01:38:37 +02:00
kijai
fbb432ce0f
Experimental TeaCache
...
I haven't figured out the whole coefficiency calculation part, however I noticed it's already super close to the time embed as it is, if we just skip the initial steps from the calculations. Seems to work okayish with the 1.3B model at least.
2025-03-02 01:34:55 +02:00
kijai
df6d0b677f
Make I2V resolution adjustment optional and other tweaks
2025-03-01 16:58:10 +02:00
kijai
16cecadf2b
make full fp16 accumulation optional
2025-03-01 12:49:14 +02:00
kijai
8d6d0ae3d8
Allow loading comfy models
2025-03-01 03:23:36 +02:00
kijai and Benjamin Paine
e5a6cc872a
Much improved context window blending
...
Co-Authored-By: Benjamin Paine <57536852+painebenjamin@users.noreply.github.com >
2025-03-01 02:52:02 +02:00
kijai
889efe7371
fix
2025-02-28 18:00:31 +02:00
kijai
b6503bc248
remove prints
2025-02-28 16:30:00 +02:00
kijai
7769725e89
fix context scheduling
2025-02-28 16:26:54 +02:00
kijai
ea162ad24e
doesn't work yet
2025-02-28 16:06:38 +02:00
kijai
7a6418a953
Fix something stupid and start implementing context schedule
2025-02-28 15:56:19 +02:00
kijai
dfe8000e63
Update nodes.py
2025-02-28 13:43:55 +02:00
kijai
a2bb63d546
Update nodes.py
2025-02-28 13:28:48 +02:00
kijai
f751818105
Apply RifleX only on temporal
2025-02-28 13:06:21 +02:00
kijai
5ee084cd01
Update nodes.py
2025-02-28 12:31:45 +02:00
kijai
076ba55b1d
Use fp16_accumulation if available when using fp16 base_precision
2025-02-28 12:22:41 +02:00
kijai
ec30087db6
Update nodes.py
2025-02-28 01:05:55 +02:00
kijai
dd3eedcd86
Add enhance-a-video, start Spargeattn testing
2025-02-28 00:49:17 +02:00
kijai
9e957121be
better preview
2025-02-27 11:13:30 +02:00
kijai
c6e6af841d
Lora block select
2025-02-27 00:07:30 +02:00
kijai
23ef23826b
Better error
2025-02-26 18:20:29 +02:00
kijai
130424077d
noise_augment to clip encode node as well
2025-02-26 16:52:42 +02:00
kijai
ab4495dcaa
Update nodes.py
2025-02-26 16:30:38 +02:00
kijai