Commit Graph

242 Commits

Author SHA1 Message Date
Ryan Dick
5fad379192 Add ability to control regional prompt region weights. 2024-03-03 12:55:07 -05:00
Ryan Dick
ad18429fe3 Very experimentation with various regional prompting tuning params. 2024-03-02 17:43:21 -05:00
Ryan Dick
942efa011e Implement (very slow) self-attention regional masking. 2024-03-01 18:43:32 -05:00
Ryan Dick
ffc4ebb14c Merge branch 'ryan/remove-attention-map-saving' into ryan/regional-conditioning 2024-03-01 11:33:50 -05:00
Ryan Dick
5b3adf0740 Remove unused code for attention map saving. 2024-02-29 23:42:37 -05:00
Ryan Dick
a5c94fba43 Delete unused functions from shared_invokeai_diffusion.py. 2024-02-29 23:41:15 -05:00
Ryan Dick
3e14bd6c45 Remove unused constructor declared with typo in name: __int__. 2024-02-29 22:42:59 -05:00
Ryan Dick
8721926f14 Merge sequential conditioning and cac conditioning logic to eliminate a bunch of duplication. 2024-02-29 22:42:59 -05:00
Ryan Dick
d87ff3a206 Remove outdated comments related to T2I-Adapters and ControlNets. 2024-02-29 22:42:59 -05:00
Ryan Dick
7d9671014b Remove use of **kwargs in do_unet_step(...), where full parameter list is known and supported. 2024-02-29 22:42:59 -05:00
Ryan Dick
4a1acd4db9 Fix avoid storing extra conditioning info in two places. 2024-02-29 22:42:59 -05:00
psychedelicious
dd9daf8efb chore: ruff 2024-03-01 10:42:33 +11:00
dunkeroni
06cc57d82a feat(nodes): added gradient mask node 2024-03-01 10:42:33 +11:00
psychedelicious
5a3195f757 final tidying before marking PR as ready for review
- Replace AnyModelLoader with ModelLoaderRegistry
- Fix type check errors in multiple files
- Remove apparently unneeded `get_model_config_enum()` method from model manager
- Remove last vestiges of old model manager
- Updated tests and documentation

resolve conflict with seamless.py
2024-03-01 10:42:33 +11:00
Lincoln Stein
a23dedd2ee make model manager v2 ready for PR review
- Replace legacy model manager service with the v2 manager.

- Update invocations to use new load interface.

- Fixed many but not all type checking errors in the invocations. Most
  were unrelated to model manager

- Updated routes. All the new routes live under the route tag
  `model_manager_v2`. To avoid confusion with the old routes,
  they have the URL prefix `/api/v2/models`. The old routes
  have been de-registered.

- Added a pytest for the loader.

- Updated documentation in contributing/MODEL_MANAGER.md
2024-03-01 10:42:33 +11:00
Lincoln Stein
78ef946e01 BREAKING CHANGES: invocations now require model key, not base/type/name
- Implement new model loader and modify invocations and embeddings

- Finish implementation loaders for all models currently supported by
  InvokeAI.

- Move lora, textual_inversion, and model patching support into
  backend/embeddings.

- Restore support for model cache statistics collection (a little ugly,
  needs work).

- Fixed up invocations that load and patch models.

- Move seamless and silencewarnings utils into better location
2024-03-01 10:42:33 +11:00
psychedelicious
05fb485d33 feat(nodes): move ConditioningFieldData to conditioning_data.py 2024-03-01 10:42:33 +11:00
Ryan Dick
8989a6cdc6 Get multi-prompt attention working simultaneously with IP-adapter. 2024-02-29 14:54:13 -05:00
Ryan Dick
1bbd4f751d Fixup logic around compatibility of prompt-to-prompt, IP-Adapter, regional prompting. 2024-02-29 12:47:23 -05:00
Ryan Dick
e132afb705 Make regional prompting work with sequential conditioning. 2024-02-28 21:21:50 -05:00
Ryan Dick
5f49e7ae26 Move regional prompt concatenation further up the stack. This solves a number of issues. 2024-02-28 20:11:47 -05:00
Ryan Dick
53ebca58ff Rename ConditioningData to TextConditioningData. 2024-02-28 13:53:56 -05:00
Ryan Dick
ee1b3157ce Split ip_adapter_conditioning out from ConditioningData. 2024-02-28 13:49:02 -05:00
Ryan Dick
e7ec13f209 Remove scheduler_args from ConditioningData structure. 2024-02-28 12:15:39 -05:00
Ryan Dick
cad3e5dbd7 Remove dead code related to an old symmetry feature. 2024-02-28 11:29:52 -05:00
Ryan Dick
845c4e93ae Update various comments related to regional prompting, and delete duplicate _preprocess_regional_prompt_mask(...) function. 2024-02-28 10:20:22 -05:00
Ryan Dick
54971afe44 Add symmetric support for regional negative text prompts. 2024-02-27 20:05:02 -05:00
Ryan Dick
cfba51aed5 Removed unused function: _prepare_text_embeddings(...) 2024-02-27 19:23:20 -05:00
Ryan Dick
2966c8de2c Handle conditioned and unconditioned text conditioning in the same way for regional prompt attention. 2024-02-27 18:16:01 -05:00
Ryan Dick
b0fcbe552e Tidy invocation interfaces for RectangleMaskInvocation and AddConditioningMaskInvocation. 2024-02-26 17:34:37 -05:00
Ryan Dick
d132fb4818 Get RegionalPromptAttnProcessor2_0 working with a ton of hacks. 2024-02-17 19:56:37 -05:00
Ryan Dick
2d5d370f38 Route masks into the RegionalPromptAttnProcessor2_0 processors. 2024-02-16 19:35:24 -05:00
Ryan Dick
caa690e24d Add concatenation of multiple text conditioning tensors, and patching of RegionalPromptAttnProcessor2_0 into the UNet. 2024-02-16 17:09:06 -05:00
Ryan Dick
ba4788007f Initialize a RegionalPromptAttnProcessor2_0 class by copying AttnProcessor2_0 from diffusers. 2024-02-15 17:52:44 -05:00
Ryan Dick
ef51005881 Remove unused code for attention map saving. 2024-02-15 17:28:55 -05:00
Ryan Dick
7b0326d7f7 Delete unused functions from shared_invokeai_diffusion.py. 2024-02-15 17:22:37 -05:00
Ryan Dick
f590b39f88 Add support for a list of ConditioningFields in DenoiseLatents. 2024-02-15 14:41:54 -05:00
Ryan Dick
382fa57f3b Remove unused constructor declared with typo in name: __int__. 2024-02-14 18:18:58 -05:00
Ryan Dick
ee3abc171d Merge sequential conditioning and cac conditioning logic to eliminate a bunch of duplication. 2024-02-14 18:17:46 -05:00
Ryan Dick
bf72cee555 Remove outdated comments related to T2I-Adapters and ControlNets. 2024-02-14 17:37:40 -05:00
Ryan Dick
e866e3b19f Remove use of **kwargs in do_unet_step(...), where full parameter list is known and supported. 2024-02-14 17:37:32 -05:00
Ryan Dick
16e574825c Fix avoid storing extra conditioning info in two places. 2024-02-14 15:34:15 -05:00
Wubbbi
73a077956b Why did my IDE change the comment? 2024-02-01 20:40:28 -05:00
Wubbbi
5e1e50bd47 Fix hopefully last import 2024-02-01 20:40:28 -05:00
Jonathan
83a9e26cd8
Respect torch-sdp in config.yaml (#5353)
If the user specifies `torch-sdp` as the attention type in `config.yaml`, we can go ahead and use it (if available) rather than always throwing an exception.
2023-12-28 05:46:28 +00:00
Ryan Dick
5127e9df2d Fix error caused by bump to diffusers 0.24. We pass kwargs now instead of positional args so that we are more robust to future changes. 2023-12-13 09:17:30 -05:00
Damian Stewart
0beb08686c
Add CFG Rescale option for supporting zero-terminal SNR models (#4335)
* add support for CFG rescale

* fix typo

* move rescale position and tweak docs

* move input position

* implement suggestions from github and discord

* cleanup unused code

* add back dropped FieldDescription

* fix(ui): revert unrelated UI changes

* chore(nodes): bump denoise_latents version 1.4.0 -> 1.5.0

* feat(nodes): add cfg_rescale_multiplier to metadata node

* feat(ui): add cfg rescale multiplier to linear UI

- add param to state
- update graph builders
- add UI under advanced
- add metadata handling & recall
- regen types

* chore: black

* fix(backend): make `StableDiffusionGeneratorPipeline._rescale_cfg()` staticmethod

This doesn't need access to class.

* feat(backend): add docstring for `_rescale_cfg()` method

* feat(ui): update cfg rescale mult translation string

---------

Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
2023-11-30 20:55:20 +11:00
psychedelicious
6494e8e551 chore: ruff format 2023-11-11 10:55:40 +11:00
psychedelicious
513fceac82 chore: ruff check - fix pycodestyle 2023-11-11 10:55:33 +11:00
psychedelicious
99a8ebe3a0 chore: ruff check - fix flake8-bugbear 2023-11-11 10:55:28 +11:00