Currently translated at 53.8% (657 of 1219 strings)
Co-authored-by: Alexander Eichhorn <pfannkuchensack@einfach-doof.de>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/de/
Translation: InvokeAI/Web UI
## What type of PR is this? (check all applicable)
- [ ] Refactor
- [ ] Feature
- [X] Bug Fix
- [ ] Optimization
- [ ] Documentation Update
- [ ] Community Node Submission
## Have you discussed this change with the InvokeAI team?
- [ ] Yes
- [X] No, because:
## Have you updated all relevant documentation?
- [ ] Yes
- [X] No
## Description
Removing LowRA from the initial models as it's been deleted from
CivitAI.
## Related Tickets & Documents
https://discord.com/channels/1020123559063990373/1168415065205112872
- Related Issue #
- Closes #
## QA Instructions, Screenshots, Recordings
<!--
Please provide steps on how to test changes, any hardware or
software specifications as well as any other pertinent information.
-->
## Added/updated tests?
- [ ] Yes
- [ ] No : _please replace this line with details on why tests
have not been included_
## [optional] Are there any post deployment tasks we need to perform?
## What type of PR is this? (check all applicable)
- [ ] Refactor
- [ ] Feature
- [ ] Bug Fix
- [x] Optimization
- [ ] Documentation Update
- [ ] Community Node Submission
## Have you discussed this change with the InvokeAI team?
- [x] Yes
- [ ] No, because:
## Have you updated all relevant documentation?
- [x] Yes
- [ ] No
## Description
Improve LoRA patching speed with the following changes:
- Calculate LoRA layer weights on the same device as the target model.
Prior to this change, weights were always calculated on the CPU. If the
target model is on the GPU, this significantly improves performance.
- Move models to their target devices _before_ applying LoRA patches.
- Improve the ordering of Tensor copy / cast operations.
## QA Instructions, Screenshots, Recordings
Tests:
- [x] Tested with a CUDA GPU, saw savings of ~10secs with 1 LoRA applied
to an SDXL model.
- [x] No regression in CPU-only environment
- [ ] No regression (and possible improvement?) on Mac with MPS.
- [x] Weights get restored correctly after using a LoRA
- [x] Stacking multiple LoRAs
Please hammer away with a variety of LoRAs in case there is some edge
case that I've missed.
## Added/updated tests?
- [x] Yes (Added some minimal unit tests. Definitely would benefit from
more, but it's a step in the right direction.)
- [ ] No
Currently translated at 51.7% (630 of 1217 strings)
Co-authored-by: Alexander Eichhorn <pfannkuchensack@einfach-doof.de>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/de/
Translation: InvokeAI/Web UI
