1e0e861b0d
Merge pull request #1604 from DenOfEquity/smol-sampler-fix
...
Update sampling.py
2024-08-30 14:49:08 +01:00
b57056b056
Update sampling.py
...
shuffle lines so DPM++2M SDE and DPM++3M SDE work with steps == 1
2024-08-30 14:42:34 +01:00
4c9380c46a
Speed up quant model loading and inference ...
...
... based on 3 evidences:
1. torch.Tensor.view on one big tensor is slightly faster than calling torch.Tensor.to on multiple small tensors.
2. but torch.Tensor.to with dtype change is significantly slower than torch.Tensor.view
3. “baking” model on GPU is significantly faster than computing on CPU when model load.
mainly influence inference of Q8_0, Q4_0/1/K and loading of all quants
2024-08-30 00:49:05 -07:00
3d62fa9598
reduce prints
2024-08-29 20:17:32 -07:00
c239200705
fix
2024-08-29 19:18:40 -07:00
95e16f7204
maintain loading related
...
1. revise model moving orders
2. less verbose printing
3. some misc minor speedups
4. some bnb related maintain
2024-08-29 19:05:48 -07:00
c70fb38b0a
Merge pull request #1581 from altoiddealer/update_readme_api
...
Update README.md
2024-08-29 20:09:54 -04:00
32cb4a1e62
Merge pull request #1579 from DenOfEquity/iterative-hrfix-quickbutton
...
allow repeated use of hrfix quickbutton
2024-08-29 21:53:55 +01:00
249446ca58
Update README.md
...
API is working normally after recent commits.
Flux usage is limited by inability to set Text Encoders.
2024-08-29 11:00:29 -04:00
2b14bcbc6e
allow repeated use of hrfix quickbutton
...
as discussed in #1535
2024-08-29 13:41:32 +01:00
948e91458a
Merge pull request #1570 from DenOfEquity/latentmodifier
...
fix issue 258 (latent modifier built-in extension)
2024-08-29 00:54:04 +01:00
ea6d03bde1
fix #258
...
fix longstanding bug with Dyn Cfg Augmentation and batch >= 2 (#258 )
2024-08-29 00:45:52 +01:00
d339600181
fix
2024-08-28 09:56:18 -07:00
11a2c0629a
Apply settings in ui-config.json to "All" presets ( #1541 )
...
* Add preset to load settings from ui-config.json
* Remove added presets and apply ui-config.json to all
2024-08-28 09:15:07 -07:00
81d8f55bca
support pytorch 2.4 new normalization features
2024-08-28 09:08:26 -07:00
0abb6c4686
Second Attempt for #1502
2024-08-28 08:08:40 -07:00
37fcb7bce8
Merge pull request #1559 from DenOfEquity/option-ignore-lorahashes-infotext
...
add 'Lora hashes' to infotext fields that can be ignored
2024-08-28 12:44:48 +01:00
408054f7e5
add 'Lora hashes' to infotext that can be ignored
...
In *Settings -> Disregard fields from pasted infotext* there is a very long list of things that can optionally be ignored when parsing infotext. Now it is a slightly longer list, and includes `Lora hashes`.
2024-08-28 11:58:32 +01:00
25662974f8
try to test #1502
2024-08-27 18:42:00 -07:00
15f41b2e5e
we do not need initial model download now
...
since we do not load model when UI start
2024-08-27 17:59:56 -07:00
20cdcf05f5
experimental: allow some tabs to hide model bar
2024-08-27 17:45:44 -07:00
db6448df74
Update Status
2024-08-27 08:34:53 -07:00
9c03493262
add some misc fix related to info text
2024-08-27 07:51:28 -07:00
d1121baf80
Adding Depth Anything v2 to ControlNet Integrated ( #1519 )
...
* Update install.py
adding install of depth anything v2
* Add files via upload
adding depth anything v2 preprocessor
* Update preprocessor_compiled.py
adding preprocessor
* Update preprocessor.py
adding preprocessor functions
2024-08-27 06:55:31 -07:00
3c4eb78cd1
Merge pull request #1531 from DenOfEquity/fix-the-hrfix-fix
...
Update txt2img.py: fix the hrfix fix
2024-08-27 00:55:42 +01:00
d7b8127438
Update txt2img.py: fix the hrfix fix
...
main issue: Upscaling would fail on single image when controlnet used.
minor issues: On the way to fixing my oversight of not accounting for control images in the gallery, I found that attempting to upscale a control image would fail due to trying to access infotext that doesn't exist. Then I handled a case previously caught by an assert more gracefully. Unhandled, these minor issues would lose the current gallery, so these extra fixes are good QoL.
Then I found another related minor issue if grids are not displayed.
2024-08-27 00:40:21 +01:00
18b356c0dc
fix
2024-08-26 16:15:07 -07:00
e94130a82a
Fix Checkpoint Merging #1359,#1095 ( #1454 )
...
- checkpoint_list[] contains the CheckpointInfo.title which is "checkpointname.safetensor [hash]"
when a checkpoint is selected to be loaded during merge, we try to match it with just "checkpointname.safetensor".
-> use checkpoint_aliases[] which already contains the checkpoint key in all possible variants.
- replaced removed sd_models.read_state_dict() with sd_models.load_torch_file()
- replaced removed sd_vae.load_vae_dict() with sd_vae.load_torch_file()
- uncommented create_config() for now, since it calls a removed method: sd_models_config.find_checkpoint_config_near_filename()
2024-08-26 15:39:37 -07:00
d55e6b5bfe
Replace API sd-vae with sd-modules ( #1463 )
2024-08-26 15:08:02 -07:00
230e39110e
Merge pull request #1492 from DenOfEquity/restore-postprocessing_focal_crop-script
...
Restore postprocessing_focal_crop script
2024-08-26 16:19:07 +01:00
37d8f87ae9
Merge pull request #1507 from DenOfEquity/cmd-arg-for-text-encoders
...
add startup argument for text encoders location
2024-08-26 15:50:21 +01:00
acf99dd74e
fix old version of pytorch
2024-08-26 06:51:48 -07:00
f22b80ef94
restrict baking to 16bits
2024-08-26 06:16:13 -07:00
7cd94babdd
Revert "Update README.md"
...
This reverts commit 891e355fc8
2024-08-26 04:52:46 -07:00
0a1fb97679
Merge branch 'lllyasviel:main' into cmd-arg-for-text-encoders
2024-08-26 11:28:00 +01:00
a0e317518e
update cmd_args.py: description
...
extra description for old clip argument
2024-08-26 11:27:35 +01:00
388b70134b
fix offline loras
2024-08-25 20:28:40 -07:00
891e355fc8
Update README.md
2024-08-25 17:38:54 -07:00
b25b62da96
fix T5 not baked
2024-08-25 17:31:50 -07:00
cae37a2725
fix dequant of unbaked parameters
2024-08-25 17:24:31 -07:00
82dfc2b15b
Significantly speed up Q4_0, Q4_1, Q4_K
...
by precomputing all possible 4bit dequant into a lookup table and use pytorch indexing to get dequant, rather than really computing the bit operations.
This should give very similar performance to native CUDA kernels, while being LoRA friendly and more flexiable
2024-08-25 16:49:33 -07:00
e60bb1c96f
Make Q4_K_S as fast as Q4_0
...
by baking the layer when model load
2024-08-25 15:02:54 -07:00
868f662eb6
fix
2024-08-25 14:44:01 -07:00
13d6f8ed90
revise GGUF by precomputing some parameters
...
rather than computing them in each diffusion iteration
2024-08-25 14:30:09 -07:00
ba01ad3711
Merge pull request #1489 from DenOfEquity/pnginfo
...
allow repeated drag-drop into PNGInfo tab
2024-08-25 16:27:39 -04:00
001f2305de
Update cmd_args.py: add text-encoder-dir
2024-08-25 15:03:02 +01:00
8c6b64e6e9
main_entry.py: add support for new text_encoder_dir cmd arg
...
copying method used for VAE
2024-08-25 15:01:30 +01:00
187faf3c5a
restore postprocessing_focal_crop.py
...
with one line fix - convert input image to RGB
2024-08-25 13:24:55 +01:00
fc9b80cf0e
restore autocrop.py
2024-08-25 13:23:21 +01:00
424baa110d
Update dragdrop.js
...
remove unnecessary check for PNGInfo element
enables repeated drag-drop into PNG Info
fixes https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/1485
2024-08-25 10:14:45 +01:00
DenOfEquity