0d8e8abd90
Default ro smaller blocks getting flipflopped first
2025-10-02 18:00:21 -07:00
d5001ed90e
Make flux support flipflop
2025-10-02 17:53:22 -07:00
8d7b22b720
Fixed FlipFlipModule.execute_blocks having hardcoded strings from Qwen
2025-10-02 17:49:43 -07:00
6d3ec9fcf3
Simplified flipflop setup by adding FlipFlopModule.execute_blocks helper
2025-10-02 16:46:37 -07:00
c4420b6a41
Change log string slightly
2025-10-02 15:34:35 -07:00
a282586995
Merge branch 'master' into flipflop-stream
2025-10-02 15:03:26 -07:00
e9364ee279
Turn on TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL by default. ( #10168 )
2025-10-02 17:57:15 -04:00
f6e3e9a456
fix(api-nodes): made logging path to be smaller ( #10156 )
2025-10-02 14:50:31 -07:00
8f4ee9984c
convert nodes_morphology.py to V3 schema ( #10159 )
2025-10-02 13:53:00 -07:00
0e9d1724be
Add a .bat to the AMD portable to disable smart memory. ( #10153 )
2025-10-02 00:33:05 -04:00
0df61b5032
Fix improper index slicing for flipflop get blocks, add extra log message
2025-10-01 21:21:36 -07:00
7c896c5567
Initial automatic support for flipflop within ModelPatcher - only Qwen Image diffusion_model uses FlipFlopModule currently
2025-10-01 20:13:50 -07:00
4965c0e2ac
WAN: Fix cache VRAM leak on error ( #10141 )
...
If this suffers an exception (such as a VRAM oom) it will leave the
encode() and decode() methods which skips the cleanup of the WAN
feature cache. The comfy node cache then ultimately keeps a reference
this object which is in turn reffing large tensors from the failed
execution.
The feature cache is currently setup at a class variable on the
encoder/decoder however, the encode and decode functions always clear
it on both entry and exit of normal execution.
Its likely the design intent is this is usable as a streaming encoder
where the input comes in batches, however the functions as they are
today don't support that.
So simplify by bringing the cache back to local variable, so that if
it does VRAM OOM the cache itself is properly garbage when the
encode()/decode() functions dissappear from the stack.
2025-10-01 18:42:16 -04:00
911331c06c
sd: fix VAE tiled fallback VRAM leak ( #10139 )
...
When the VAE catches this VRAM OOM, it launches the fallback logic
straight from the exception context.
Python however refs the entire call stack that caused the exception
including any local variables for the sake of exception report and
debugging. In the case of tensors, this can hold on the references
to GBs of VRAM and inhibit the VRAM allocated from freeing them.
So dump the except context completely before going back to the VAE
via the tiler by getting out of the except block with nothing but
a flag.
The greately increases the reliability of the tiler fallback,
especially on low VRAM cards, as with the bug, if the leak randomly
leaked more than the headroom needed for a single tile, the tiler
would fallback would OOM and fail the flow.
2025-10-01 18:40:28 -04:00
bb32d4ec31
feat: Add Epsilon Scaling node for exposure bias correction ( #10132 )
2025-10-01 17:59:07 -04:00
a6f83a4a1a
Support the new hunyuan vae. ( #10150 )
2025-10-01 17:19:13 -04:00
e4f99b479a
convert nodes_ip2p.pt to V3 schema ( #10097 )
2025-10-01 12:20:30 -07:00
d9c0a4053d
convert nodes_lt.py to V3 schema ( #10084 )
2025-10-01 12:19:56 -07:00
11bab7be76
convert nodes_pag.py to V3 schema ( #10080 )
2025-10-01 12:18:49 -07:00
3af1881455
convert nodes_optimalsteps.py to V3 schema ( #10074 )
2025-10-01 12:18:04 -07:00
e0210ce0a7
convert nodes_differential_diffusion.py to V3 schema ( #10056 )
2025-10-01 12:17:33 -07:00
7eb7160db4
convert nodes_gits.py to V3 schema ( #9949 )
2025-10-01 12:16:59 -07:00
ec156e72eb
Merge branch 'master' into flipflop-stream
2025-09-30 23:08:37 -07:00
01f4512bf8
In-progress commit on making flipflop async weight streaming native, made loaded partially/loaded completely log messages have labels because having to memorize their meaning for dev work is annoying
2025-09-30 23:08:08 -07:00
638097829d
convert nodes_audio_encoder.py to V3 schema ( #10123 )
2025-09-30 23:00:22 -07:00
c4a8cf60ab
Bump frontend to 1.27.7 ( #10133 )
2025-09-30 22:12:32 -07:00
bab8ba20bf
ComfyUI version 0.3.62.
2025-09-30 15:12:07 -04:00
b682a73c55
enable Seedance Pro model in the FirstLastFrame node ( #10120 )
2025-09-30 10:43:41 -07:00
631b9ae861
fix(Rodin3D-Gen2): missing "task_uuid" parameter ( #10128 )
2025-09-30 10:21:47 -07:00
f48d7230de
Add new portable links to readme. ( #10112 )
2025-09-30 12:17:49 -04:00
d0bd221495
Merge branch 'master' into flipflop-stream
2025-09-29 22:49:38 -07:00
8a8162e8da
Fix percentage logic, begin adding elements to ModelPatcher to track flip flop compatibility
2025-09-29 22:49:12 -07:00
6e079abc3a
Workflow permission fix. ( #10110 )
2025-09-29 23:11:37 -04:00
977a4ed8c5
ComfyUI version 0.3.61
2025-09-29 23:04:42 -04:00
414a178fb6
Add basic readme for AMD portable. ( #10109 )
2025-09-29 23:03:02 -04:00
447884b657
Make stable release workflow callable. ( #10108 )
2025-09-29 20:37:51 -04:00
bed4b49d08
Add action to do the full stable release. ( #10107 )
2025-09-29 20:31:15 -04:00
342cf644ce
Add a way to have different names for stable nvidia portables. ( #10106 )
2025-09-29 20:05:44 -04:00
3758848423
Different base files for nvidia and amd portables. ( #10105 )
2025-09-29 19:54:37 -04:00
0db6aabed3
Different base files for different release. ( #10104 )
2025-09-29 19:54:05 -04:00
ff789c8beb
Merge branch 'master' into flipflop-stream
2025-09-29 16:09:51 -07:00
1673ace19b
Make the final release test optional in the stable release action. ( #10103 )
2025-09-29 19:08:42 -04:00
7f38e4c538
Add action to create cached deps with manually specified torch. ( #10102 )
2025-09-29 17:27:52 -04:00
8accf50908
convert nodes_mahiro.py to V3 schema ( #10070 )
2025-09-29 12:35:51 -07:00
ed0f4a609b
dont cache new locale entry points ( #10101 )
2025-09-29 12:16:02 -07:00
041b8824f5
convert nodes_perpneg.py to V3 schema ( #10081 )
2025-09-29 12:05:28 -07:00
b1111c2062
convert nodes_mochi.py to V3 schema ( #10069 )
2025-09-29 12:03:35 -07:00
05a258efd8
add WanImageToImageApi node ( #10094 )
2025-09-29 12:01:04 -07:00
c8276f8c6b
Update template to 0.1.91 ( #10096 )
2025-09-29 11:59:42 -07:00
6ec1cfe101
[Rodin3d api nodes] Updated the name of the save file path (changed from timestamp to UUID). ( #10011 )
...
* Update savepath name from time to uuid
* delete lib
2025-09-29 11:59:12 -07:00
Jedrzej Kosinski
comfyanonymous
Alexander Piskun
rattus128
Koratahiu
AustinMroz
comfyanonymous
Christian Byrne
ComfyUI Wiki
Changrz