Commit Graph

2540 Commits

Author SHA1 Message Date
fc3378bb74 Load legacy ckpt files as diffusers models (#2468)
* refactor ckpt_to_diffuser to allow converted pipeline to remain in memory

- This idea was introduced by Damian
- Note that although I attempted to use the updated HuggingFace module
  pipelines/stable_diffusion/convert_from_ckpt.py, it was unable to
  convert safetensors files for reasons I didn't dig into.
- Default is to extract EMA weights.

* add --ckpt_convert option to load legacy ckpt files as diffusers models

- not quite working - I'm getting artifacts and glitches in the
  converted diffuser models
- leave as draft for time being

* do not include safety checker in converted files

* add ability to control which vae is used

API now allows the caller to pass an external VAE model to the
checkpoint conversion process. In this way, if an external VAE is
specified in the checkpoint's config stanza, this VAE will be used
when constructing the diffusers model.

Tested with both regular and inpainting 1.X models.

Not tested with SD 2.X models!

---------

Co-authored-by: Jonathan <34005131+JPPhoto@users.noreply.github.com>
Co-authored-by: Damian Stewart <null@damianstewart.com>
2023-02-02 20:15:44 +00:00
1fe5ec32f5 Swap codeowners for installer (#2477)
This PR changes the codeowner for the installer directory from
@tildebyte to @ebr due to the former's time commitments.

Further reorganization of the codeowners is pending.
2023-02-02 14:27:31 -05:00
6dee9051a1 swap codeowners for installer 2023-02-02 13:54:53 -05:00
80c5322ccc fix(img2img): do not attempt to do a zero-step img2img when strength is low (#2472) 2023-02-02 10:04:09 -08:00
da181ce64e Merge branch 'main' into fix/img2img-low-strength 2023-02-02 09:40:16 -08:00
5ef66ca237 Fix typo in xformers version, 0.16 -> 0.0.16 (#2475) 2023-02-02 09:39:08 -08:00
7aa331af8c Merge branch 'main' into fix/img2img-low-strength 2023-02-02 07:20:34 -08:00
9e943ff7dc Fix typo in xformers version, 0.16 -> 0.0.16 2023-02-02 05:26:15 -05:00
b5040ba8d0 Build 2023-02-02 22:52:03 +13:00
07462d1d99 Remove Inpaint Replace 2023-02-02 22:52:03 +13:00
44d36a0e0b fix(img2img): do not attempt to do a zero-step img2img when strength is low 2023-02-01 18:42:54 -08:00
31146eb797 add workflow to clean caches after PR gets closed (#2450)
This helps at least a bit to get rid of all those huge caches
2023-02-01 19:06:07 +01:00
99cd598334 add workflow to clean caches after PR gets closed 2023-02-01 18:32:29 +01:00
5441be8169 requirements: add xformers for CUDA platforms (#2465)
[xformers
0.16](https://github.com/facebookresearch/xformers/releases/tag/v0.0.16)
was released earlier today, and is now installable from wheels on PyPI!

Fixes #1876.
2023-02-01 10:59:13 -05:00
3e98b50b62 Merge branch 'main' into req-xformers 2023-02-01 10:29:49 -05:00
5f16148dea Prevent actions from running on draft PRs (#2457)
Draft PRs are triggering actions on every commit (except
`test-invoke-pip.yml`).

I've added a conditional to each job to only run when the PR is not a
draft.

(maybe there is a reason we are running all applicable workflows on
draft PRs?)
2023-02-01 00:33:15 -05:00
9628d45a92 Merge branch 'main' into build/no-actions-on-draft 2023-02-01 00:15:30 -05:00
6608343455 [enhancement] Print status message at startup when xformers is available (#2461) 2023-01-31 19:11:17 -08:00
abd972f099 Merge branch 'main' into feat/xformers-startup-message 2023-01-31 18:48:09 -08:00
bd57793a65 fix img2img by working around pytorch bug (#2458)
horribly, temporarily send the vae to `.cpu()` so that good latents can
be produced

closes #2418
2023-01-31 21:46:05 -05:00
8cdc65effc Merge branch 'main' into fix_2418_simplified 2023-01-31 17:45:54 -08:00
85b553c567 requirements: add xformers for CUDA platforms
Now available from pip!
2023-01-31 16:51:43 -08:00
af74a2d1f4 fix broken Dockerfile (#2445)
also switch to `python:3.9-slim` since it has a ton less security issues
2023-02-01 01:47:25 +01:00
6fdc9ac224 re-enable INVOKE_MODEL_RECONFIGURE 2023-02-01 01:21:07 +01:00
8107d354d9 fix broken Dockerfile
also switch to python 3.9:slim since it has a ton less security issues
2023-02-01 01:21:07 +01:00
28c17613c4 feat(inpaint): add solid infill for use with inpainting model (#2441)
A new infill method, **solid:** solid color. currently using middle
gray.

Fixes #2417

It seems like the runwayml inpainting model specifically expects those
masked areas to be blanked out like this.

I haven't tried the SD 2.0 inpainting model with it yet.
2023-01-31 18:27:48 -05:00
e6d52d7ce6 Merge branch 'main' into fix_2418_simplified 2023-01-31 18:11:56 -05:00
8c726d3e3e Merge branch 'main' into build/no-actions-on-draft 2023-01-31 18:08:52 -05:00
56e2d22b6e Merge branch 'main' into feat/solid-infill 2023-01-31 18:02:17 -05:00
053d11fe30 fix(inpainting model): blank areas to be repainted in the masked image (#2447)
Otherwise the model seems too reluctant to change these areas, even
though the mask channel should allow it to.

This makes the solid infill method proposed by #2441 less necessary,
though I think there's still a place for an infill method that is faster
than patchmatch and more predictable than tiles.

Even with #2441, this PR is still useful because it influences all areas
to be painted, not just the infill area.

Fixes #2417
2023-01-31 18:01:33 -05:00
0066187651 Merge branch 'main' into feat/solid-infill 2023-01-31 17:53:09 -05:00
d3d24fa816 fill color is parameterized 2023-01-31 17:52:33 -05:00
4d58fed6b0 Merge branch 'main' into fix/inpainting-blank-slate 2023-01-31 11:04:56 -08:00
bde5874707 fix dimension errors when inpainting model is used with hires-fix (#2440) 2023-01-31 11:04:02 -08:00
eed802f5d9 Merge branch 'main' into fix/hires_inpaint 2023-01-31 09:34:29 -08:00
fc8e3dbcd3 fix crash when editing name of model
- fixes a spurious "unknown model name" error when trying to edit the
  short name of an existing model.
- relaxes naming requirements to include the ':' and '/' characters
  in model names
2023-01-31 09:59:58 -05:00
b9cc56593e print status message at startup when xformers is available 2023-01-30 22:01:06 -05:00
6a0e1c8673 Merge branch 'main' into build/no-actions-on-draft 2023-01-31 12:00:38 +11:00
371edc993a Implement .swap() against diffusers 0.12 (#2385) 2023-01-30 15:56:24 -08:00
5299324321 workaround for pytorch bug, fixes #2418 2023-01-30 18:45:53 +01:00
817e36f8bf Merge branch 'diffusers_cross_attention_control_reimplementation' of github.com:damian0815/InvokeAI into diffusers_cross_attention_control_reimplementation 2023-01-30 16:23:52 +01:00
d044d4c577 rename override/restore methods to better reflect what they actually do 2023-01-30 16:23:44 +01:00
3f1120e6f2 Merge branch 'main' into diffusers_cross_attention_control_reimplementation 2023-01-30 16:17:25 +01:00
17d73d09c0 Revert "with diffusers cac, always run the original prompt on the first step"
This reverts commit 27ee939e4b.
2023-01-30 15:38:03 +01:00
478c379534 for cac make t_start=0.1 the default 2023-01-30 15:30:01 +01:00
c5c160a788 Merge branch 'diffusers_cross_attention_control_reimplementation' of github.com:damian0815/InvokeAI into diffusers_cross_attention_control_reimplementation 2023-01-30 14:51:06 +01:00
27ee939e4b with diffusers cac, always run the original prompt on the first step 2023-01-30 14:50:57 +01:00
c222cf7e64 Prevents actions from running on draft PRs 2023-01-30 22:28:05 +11:00
d9ed0f6005 fix documentation of huggingface cache location (#2430)
* fix documentation of huggingface cache location

---------

Co-authored-by: Jonathan <34005131+JPPhoto@users.noreply.github.com>
2023-01-29 20:30:50 -06:00
5ce62e00c9 Merge branch 'main' into diffusers_cross_attention_control_reimplementation 2023-01-29 13:52:01 -06:00