fba40eb1bd
Fix the padding behavior when max-pooling regional prompt masks to mirror the downscaling behavior of SD and SDXL. Prior to this change, denoising with input latent dimensions that were not evenly divisible by 8 would raise an exception.
2024-04-09 16:50:43 -04:00
0c6dd32ece
(minor) Fix IP-Adapter conditional logic in CustomAttnProcessor2_0.
2024-04-09 15:06:51 -04:00
0bdbfd4d1d
Add support for IP-Adapter masks.
2024-04-09 15:06:51 -04:00
2e27ed5f3d
Pass IP-Adapter scales through the cross_attn_kwargs pathway, since they are the same for all attention layers. This change also helps to prepare for adding IP-Adapter region masks.
2024-04-09 15:06:51 -04:00
4a828818da
Remove support for Prompt-to-Prompt cross-attention control (aka .swap()). This feature is not widely used. It does not work with SDXL and is incompatible with IP-Adapter and regional prompting. The implementation is also intertwined with both text embedding and the UNet attention layers, resulting in a high maintenance burden. For all of these reasons, we have decided to drop support.
2024-04-09 10:57:02 -04:00
182810337c
Add utility to_standard_float_mask(...) to convert various mask formats to a standardized format.
2024-04-09 08:12:12 -04:00
75ef473748
Pull the upstream changes from diffusers' AttnProcessor2_0 into CustomAttnProcessor2_0. This fixes a bug in CustomAttnProcessor2_0 that was being triggered when peft was not installed. The bug was present in a block of code that was previously copied from diffusers. The bug seems to have been introduced during diffusers' migration to PEFT for their LoRA handling. The upstream bug was fixed in 531e719163.
2024-04-09 08:12:12 -04:00
a78df8123f
Update the diffusion logic to use the new regional prompting feature.
2024-04-09 08:12:12 -04:00
7ca677578e
Create a UNetAttentionPatcher for patching UNet models with CustomAttnProcessor2_0 modules.
2024-04-09 08:12:12 -04:00
31c456c1e6
Update CustomAttention to support both IP-Adapters and regional prompting.
2024-04-09 08:12:12 -04:00
2ce79b61f5
Initialize a RegionalPromptAttnProcessor2_0 class by copying AttnProcessor2_0 from diffusers.
2024-04-09 08:12:12 -04:00
109e3f0e7f
Add RegionalPromptData class for managing prompt region masks.
2024-04-09 08:12:12 -04:00
d1e45585d0
Add TextConditioningRegions to the TextConditioningData data structure.
2024-04-09 08:12:12 -04:00
aba023e0c5
Improve documentation of conditioning_data.py.
2024-04-09 08:12:12 -04:00
e354c29b52
Rename ConditioningData -> TextConditioningData.
2024-04-09 08:12:12 -04:00
a7f363e654
Split ip_adapter_conditioning out from ConditioningData.
2024-04-09 08:12:12 -04:00
9b2162e564
Remove scheduler_args from ConditioningData structure.
2024-04-09 08:12:12 -04:00
fd1f240853
fix: SDXL Refiner not working properly with Inpainting
2024-04-09 14:13:10 +10:00
9ab6655491
feat(backend): clean up choose_precision
...
- Allow user-defined precision on MPS.
- Use more explicit logic to handle all possible cases.
- Add comments.
- Remove the app_config args (they were effectively unused, just get the config using the singleton getter util)
2024-04-07 09:41:05 -04:00
4068e817d6
fix(mm): typing issues in model cache
2024-04-06 14:35:36 +11:00
a09d705e4c
fix(mm): remove vram check
...
This check prematurely reports insufficient VRAM on Windows. See #6106 for details.
2024-04-06 14:35:36 +11:00
4571986c63
fix misplaced lock call
2024-04-05 14:32:18 +11:00
812f10730f
adjust free vram calculation for models that will be removed by lazy offloading ( #6150 )
...
Co-authored-by: Lincoln Stein <lstein@gmail.com >
2024-04-04 22:51:12 -04:00
f0b1bb0327
feat(nodes): redo tile infill
...
The previous algorithm errored if the image wasn't divisible by the tile size. I've reimplemented it from scratch to mitigate this issue.
The new algorithm is simpler. We create a pool of tiles, then use them to create an image composed completely of tiles. If there is any awkwardly sized space on the edge of the image, the tiles are cropped to fit.
Finally, paste the original image over the tile image.
I've added a jupyter notebook to do a smoke test of infilling methods, and 10 test images.
The other infill algorithms can be easily tested with the notebook on the same images, though I didn't set that up yet.
Tested and confirmed this gives results just as good as the earlier infill, though of course they aren't the same due to the change in the algorithm.
2024-04-05 08:49:13 +11:00
3c195d74a5
fix: bypass edge pixels which cannot transform to tile size
...
Still need to fix this somehow
2024-04-05 08:49:13 +11:00
32a6b758cd
wip: Initial Infill Methods Refactor
2024-04-05 08:49:13 +11:00
51ca59c088
Update probe to always use cpu for loading models
2024-04-04 07:34:43 +11:00
85f53f94f8
feat(mm): include needed vs free in OOM
...
Gives us a bit more visibility into these errors, which seem to be popping up more frequently with the new MM.
2024-04-04 06:26:15 +11:00
be574cb764
fix: incorrect suffix check in ip adapter checkpoint file
2024-04-03 22:38:28 +05:30
5f01de1993
chore: ruff and lint fixes
2024-04-03 20:41:51 +05:30
e574815413
chore: clean up merge conflicts
2024-04-03 20:28:00 +05:30
fb293dcd84
Merge branch 'checkpoint-ip-adapter' of https://github.com/blessedcoolant/InvokeAI into checkpoint-ip-adapter
2024-04-03 20:23:07 +05:30
414851f2f0
fix: raise and present the runtime error from the exception
2024-04-03 20:21:50 +05:30
2dcbb7223b
fix: use Path for ip_adapter_ckpt_path instead of str
2024-04-03 20:21:03 +05:30
14a9f74b17
cleanup: use load_file of safetensors directly for loading ip adapters
2024-04-03 12:40:13 +05:30
1372ef15b3
fix: Fail when unexpected keys are found in IP Adapter models
2024-04-03 12:40:11 +05:30
be1212de9a
fix: Raise a better error when incorrect CLIP Vision model is used
2024-04-03 12:40:10 +05:30
a14ce0edab
chore: rename IPAdapterDiffusersConfig to IPAdapterInvokeAIConfig
2024-04-03 12:40:10 +05:30
4a0dfc3b2d
ui: improve the clip vision model picker layout
2024-04-03 12:40:08 +05:30
936b99bd3c
chore: improve types in ip_adapter backend file
2024-04-03 12:40:02 +05:30
79f7b61dfe
fix: cleanup across various ip adapter files
2024-04-03 12:39:52 +05:30
b1c8266e22
feat: add base model recognition for ip adapter safetensor files
2024-04-03 12:39:52 +05:30
67afb1763e
wip: Initial implementation of safetensor support for IP Adapter
2024-04-03 12:39:52 +05:30
0259114d9c
Merge branch 'main' into main
2024-04-03 17:03:19 +11:00
51e515b925
tidy: use lowercase for tcd scheduler identifier
2024-04-03 17:03:02 +11:00
8c509295f9
chore: ruff
2024-04-03 17:02:45 +11:00
59b4a23479
feat(mm): use same pattern for vae converter as others
...
Add `dump_path` arg to the converter function & save the model to disk inside the conversion function. This is the same pattern as in the other conversion functions.
2024-04-01 12:34:49 +11:00
13f410478a
fix(mm): typing issues in vae loader
2024-04-01 12:34:49 +11:00
25ff0bf80f
fix(mm): return converted vae model instead of path
...
This was missed in #6072 .
2024-04-01 12:34:49 +11:00
23390f1516
cleanup: use load_file of safetensors directly for loading ip adapters
2024-04-01 06:37:38 +05:30
