maybleMyers
5612bc2ccd
remove debug
2025-08-30 04:39:48 -07:00
maybleMyers
6862ac07c5
smarter offload
2025-08-30 04:37:38 -07:00
maybleMyers
0418f18a75
offloading strategy test
2025-08-30 04:07:38 -07:00
maybleMyers
4a8a88212c
ops
2025-08-30 03:20:09 -07:00
maybleMyers
cc42c709c4
update memory calculation for radiance models
2025-08-30 02:32:18 -07:00
maybleMyers
385c89a294
remove bugs
2025-08-28 16:54:10 -07:00
maybleMyers
0de6a74b15
update vram management
2025-08-28 16:07:49 -07:00
maybleMyers
388cacde82
testing things
2025-08-28 08:50:58 -07:00
maybleMyers
9464505b45
fp32 option for te
2025-06-29 08:44:12 -07:00
maybleMyers
9a866bcf11
merge stuff
2025-06-29 07:19:08 -07:00
maybleMyers
fd3a3b856a
fix sage attention implementation
2025-05-27 02:01:28 -07:00
maybleMyers
f362ec6638
merge experimental sage pr
2025-05-25 18:23:12 -07:00
spawner
df925305e8
flash and sage
2025-05-19 15:34:50 +08:00
spawner
d88f325f32
sage and flash
2025-05-19 15:33:33 +08:00
spawner
4f67034794
Update args.py
2025-05-19 15:33:03 +08:00
Mathieu Croquelois
17a42e5877
Add BF16 to GGUF ( #2877 )
2025-05-19 00:06:23 -04:00
maybleMyers
bd2de22508
update to latest version of forge, fix que
2025-05-18 14:36:46 -07:00
machina
0ced1d0cd0
Support wd_on_output for DoRA ( #2856 )
2025-05-05 09:32:33 -04:00
DenOfEquity
5e1dcd35a8
some extra lora support, inc. new glora ( #2715 )
...
* support new glora (via ComfyUI)
* support BFL FluxTools loras (mostly via ComfyUI)
* also support using loras (like Hyper, Turbo) with FluxTools models
2025-03-04 00:26:43 +00:00
DenOfEquity
4f825bc070
fix zero'd cond_t5 device ( #2701 )
...
could be created on the wrong device if user has enough VRAM for Forge to run text encoders on GPU.
just move it to the cond_l device
2025-02-27 22:01:35 +00:00
DenOfEquity
3ce373939e
this is useful (spiece.model for SD3 T5 tokenizer)( #2699 )
2025-02-27 20:40:19 +00:00
DenOfEquity
f23bc80d2f
SD3+ ( #2688 )
...
Co-authored-by: graemeniedermayer graemeniedermayer@users.noreply.github.com
2025-02-27 17:54:44 +00:00
DenOfEquity
8dd92501e6
Add SDXL refiner model ( #2686 )
...
add sdxlrefiner
adjust some Settings
custom CLIP-G support
2025-02-25 10:49:47 +00:00
DenOfEquity
184bb04f8d
increased support for custom CLIPs ( #2642 )
...
increased support for custom CLIPs
more forms recognised
now can be applied to sd1.5, sdxl, (sd3)
2025-02-21 12:01:39 +00:00
DenOfEquity
4a30c15769
update for Save Checkpoint button ( #2636 )
...
function in Checkpoint Merger previously produced unusable checkpoints for non-Flux architectures because keys had names not recognised by the model loader
2025-02-09 13:21:15 +00:00
hako-mikan
daee4c0d8f
Add force refresh to LoRA Loader refresh function ( #2584 )
2025-01-28 16:04:44 -05:00
DenOfEquity
2051c4100b
Apply emphasis mode changes immediately ( #2423 )
...
Previously emphasis mode changes applied only on model load.
Also added handling for mode 'None' (literal interpretation), which previously gave results identical to mode 'Ignore'.
2024-12-08 19:25:39 +00:00
DenOfEquity
19a9a78c9b
fix/workaround for potential memory leak ( #2315 )
...
unload old models, based on reference count <= 2
(in practise only noticed extra copies of JointTextEncoder, never KModel or IntegratedAutoencoderKL)
#2281 #2308 and others
2024-11-14 22:05:54 +00:00
catboxanon
6e1a7908b4
Fallback to estimated prediction if .yaml prediction not suitable ( #2273 )
2024-11-06 23:49:18 -05:00
catboxanon
f4afbaff45
Only use .yaml config prediction if actually set ( #2272 )
2024-11-06 23:44:12 -05:00
catboxanon
05b01da01f
Fix ldm .yaml loading typo ( #2248 )
2024-11-02 10:29:13 -04:00
catboxanon
90a6970fb7
Compatibility for ldm .yaml configs ( #2247 )
2024-11-02 10:16:51 -04:00
layerdiffusion
d50f390c7e
rewrite mu shift
2024-10-31 22:57:26 -07:00
catboxanon
6f4350d65f
Fix typo in .yaml config load ( #2228 )
2024-10-31 07:09:27 -04:00
catboxanon
b691b1e755
Fix .yaml config loading ( #2224 )
2024-10-30 16:18:44 -04:00
Symbiomatrix
41a21f66fd
Restore embedding filepath. ( #2030 )
2024-10-26 23:53:48 +01:00
catboxanon
edeb2b883f
Fix loading diffusers format VAEs ( #2171 )
2024-10-24 14:27:57 -04:00
catboxanon
edc46380cc
Automatically enable ztSNR when applicable ( #2122 )
2024-10-19 20:33:34 -04:00
catboxanon
aba35cde5f
Fix Zero Terminal SNR option
2024-10-19 12:28:14 -04:00
catboxanon
f620f55e56
Fallback to already detected prediction type if none applicable found
2024-10-19 07:47:56 -04:00
catboxanon
5ec47a6b93
Fix model prediction detection
...
Closes #1109
2024-10-19 06:18:02 -04:00
DenOfEquity
cc378589a4
(T5) pad chunks to length of largest chunk ( #1990 )
...
When the prompt is chunked using the BREAK keyword, chunks will be padded to the minimum size of 256 tokens - but chunks can be longer. torch.stack then fails if all chunks are not the same size, so find the largest and pad all to match.
#1988 (doesn't quite ID the real issue, prompts longer than 255 tokens work fine)
2024-10-07 11:37:06 +01:00
DenOfEquity
2467c88c50
fix for XPU ( #1997 )
...
use float32 for XPU, same as previous fix for MPS
2024-10-06 14:33:47 +01:00
Conor Nash
8bd7e0568f
Get Flux working on Apple Silicon ( #1264 )
...
Co-authored-by: Conor Nash <conor@nbs.consulting>
2024-09-13 15:40:11 +01:00
Panchovix
c13b26ba27
Rephrase low GPU warning ( #1761 )
...
Make emphasis that it's performance degradation, and for this diffusion process.
2024-09-09 14:30:22 -03:00
layerdiffusion
efe6fed499
add a way to exchange variables between modules
2024-09-08 20:22:04 -07:00
layerdiffusion
f40930c55b
fix
2024-09-08 17:24:53 -07:00
layerdiffusion
44eb4ea837
Support T5&Clip Text Encoder LoRA from OneTrainer
...
requested by #1727
and some cleanups/licenses
PS: LoRA request must give download URL to at least one LoRA
2024-09-08 01:39:29 -07:00
layerdiffusion
a8a81d3d77
fix offline quant lora precision
2024-08-31 13:12:23 -07:00
layerdiffusion
79b25a8235
move codes
2024-08-31 11:31:02 -07:00
