86f705bf48
Optimize weights handling
2024-07-30 03:39:01 +03:00
1fd9631f2d
Comments fix
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-30 00:39:50 +03:00
2227a2357f
Suggested changes + simplify weights logic in patching
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-30 00:34:37 +03:00
8500bac3ca
Use logger for warning
2024-07-28 22:51:52 +03:00
9e582563eb
Suggested changes
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-27 04:25:15 +03:00
faa88f72bf
Make lora as separate extensions
2024-07-27 02:39:53 +03:00
46c632e7cc
Change layer detection keys according to LyCORIS repository
2024-07-25 02:10:47 +03:00
653f63ae71
Add layer keys check
2024-07-25 02:03:08 +03:00
8a9e2f57a4
Handle bias in full/diff lora layer
2024-07-25 02:02:37 +03:00
31949ed2f2
Refactor code a bit
2024-07-25 02:00:30 +03:00
0ccb304b8b
Ruff format
2024-07-24 16:01:29 +03:00
ab0bfa709a
Handle loras in modular denoise
2024-07-24 05:07:29 +03:00
e2e47fd606
Merge branch 'main' into stalker-modular_controlnet
2024-07-23 10:19:12 -04:00
db52f5606f
Merge branch 'main' into stalker-modular_freeu
2024-07-23 09:53:32 -04:00
39e804d0f8
Use consistent param names in patch_extension(...) functions: context -> ctx.
2024-07-23 09:18:04 -04:00
4e8dcb7a1a
Suggested changes
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-23 01:46:29 +03:00
3cb13d6288
Rename as suggested in other PRs
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-23 01:01:18 +03:00
5f0fe3c8a9
Suggested changes
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-22 23:09:11 +03:00
1b359b55cb
Suggested changes
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-22 22:17:29 +03:00
42356ec866
Add ControlNet support to denoise
2024-07-21 20:01:30 +03:00
1748848b7b
Ruff fixes
2024-07-21 18:37:20 +03:00
5772965f09
Fix slightly different output with old backend
2024-07-21 18:31:30 +03:00
e046e60e1c
Add FreeU support to denoise
2024-07-21 18:31:10 +03:00
9a1420280e
Add rescale cfg support to denoise
2024-07-21 17:33:43 +03:00
930ff559e4
add sdxl tile to starter models
2024-07-19 16:49:33 -04:00
78d2b1b650
Merge branch 'main' into stalker-backend_base
2024-07-19 16:25:20 -04:00
39e10d894c
Add invocation cancellation logic to patchers
2024-07-19 23:17:01 +03:00
e16faa6370
Add gradient blending to tile seams in MultiDiffusion.
2024-07-19 13:05:50 -07:00
83a86abce2
Add unit tests for ExtensionsManager and ExtensionBase.
2024-07-19 14:15:46 -04:00
0c56d4a581
Ryan's suggested changes to extension manager/extensions
...
Co-Authored-By: Ryan Dick <14897797+RyanJDick@users.noreply.github.com >
2024-07-18 23:49:44 +03:00
97a7f51721
don't use cpu state_dict for model unpatching when executing on cpu ( #6631 )
...
Co-authored-by: Lincoln Stein <lstein@gmail.com >
2024-07-18 15:34:01 -04:00
710dc6b487
Merge branch 'main' into stalker7779/backend_base
2024-07-18 01:08:04 +03:00
2ef3b49a79
Add run cancelling logic to extension manager
2024-07-17 04:39:15 +03:00
3f79467f7b
Ruff format
2024-07-17 04:24:45 +03:00
2c2ec8f0bc
Comments, a bit refactor
2024-07-17 04:20:31 +03:00
79e35bd0d3
Minor fixes
2024-07-17 03:48:37 +03:00
137202b77c
Remove patch_unet logic for now
2024-07-17 03:40:27 +03:00
03e22c257b
Convert conditioning_mode to enum
2024-07-17 03:37:11 +03:00
ae6d4fbc78
Move out _concat_conditionings_for_batch submethods
2024-07-17 03:31:26 +03:00
cd1bc1595a
Rename sequential as private variable
2024-07-17 03:24:11 +03:00
f866b49255
Add some ESRGAN and SwinIR upscale models to the starter models list.
2024-07-16 15:55:10 -04:00
b7c6c63005
Added some comments
2024-07-16 22:52:44 +03:00
6b0ca88177
Merge branch 'main' into ryan/spandrel-upscale-tiling
2024-07-16 15:40:14 -04:00
81991e072b
Merge branch 'main' into ryan/spandrel-upscale
2024-07-16 15:14:08 -04:00
cec345cb5c
Change attention processor apply logic
2024-07-16 20:03:29 +03:00
608cbe3f5c
Separate inputs in denoise context
2024-07-16 19:30:29 +03:00
38343917f8
fix(backend): revert non-blocking device transfer
...
In #6490 we enabled non-blocking torch device transfers throughout the model manager's memory management code. When using this torch feature, torch attempts to wait until the tensor transfer has completed before allowing any access to the tensor. Theoretically, that should make this a safe feature to use.
This provides a small performance improvement but causes race conditions in some situations. Specific platforms/systems are affected, and complicated data dependencies can make this unsafe.
- Intermittent black images on MPS devices - reported on discord and #6545 , fixed with special handling in #6549 .
- Intermittent OOMs and black images on a P4000 GPU on Windows - reported in #6613 , fixed in this commit.
On my system, I haven't experience any issues with generation, but targeted testing of non-blocking ops did expose a race condition when moving tensors from CUDA to CPU.
One workaround is to use torch streams with manual sync points. Our application logic is complicated enough that this would be a lot of work and feels ripe for edge cases and missed spots.
Much safer is to fully revert non-locking - which is what this change does.
2024-07-16 08:59:42 +10:00
9f088d1bf5
Multiple small fixes
2024-07-16 00:51:25 +03:00
fd8d1c12d4
Remove 'del' operator overload
2024-07-16 00:43:32 +03:00
d623bd429b
Fix condtionings logic
2024-07-16 00:31:56 +03:00
