5eac99d3f5
optimize diffusers memory handling
2023-08-18 20:41:34 +00:00
20ba9aa649
update lint
2023-08-15 08:29:04 +02:00
ebc1243238
Merge pull request #2014 from midcoastal/Issue1563/FS-Path-Cache
...
Add a FS directory/path cacher
2023-08-18 07:45:06 +02:00
67f369ed25
A walk() optimization and lint fixes
2023-08-17 16:49:42 -04:00
89da66207a
I guess I didn't quite understand Lint error C416
...
Seems like a rather pedantic check, if you ask me... Oh well.
2023-08-16 00:25:27 -04:00
03363ce866
Resolve Lint Errors: C416 (x2), W291 and F541
2023-08-15 23:20:00 -04:00
c9bdd0344d
Allow JPEG image files
...
JPEG is allowed and searched for when looking for previews/thumbs, and therefore should be allowed to view.
2023-08-15 23:10:30 -04:00
873e864640
Ooof, modelloader changes needed for TI upgrade
2023-08-15 23:09:24 -04:00
6eae768fb0
Gather HTML parts as list to append in the end
...
Somehow, significant increase in speed (size of text append was slowing large loads down.
2023-08-15 23:08:26 -04:00
05850c2344
Upgrade Lora/TI model listers to use cache
2023-08-15 23:06:23 -04:00
7e4aed7948
Add a FS directory/path cacher
...
- Added FS dir/path mechanism to 'modules.modelloader'
- Refactored 'modules.modelloader.load_models' to use the cache
- Refactored 'modules.ui_extra_networks.find_*' methods to use the cache
- Added progress indicator to 'modules.ui_extra_networks.create_html'
The cache, as implimented, will always ensure it is up-to-date (per 'directory_has_changed') and significantly improves loading speed (when used with 'model_loader' and the 'find_*' methods) with large model directories.
Overall load speed tested with ~10k models (mix of checkpoints, loras, lycoris, and embeddings) and ~40k secndary files (images, descriptions, CivitAI Info, etc). Loading speeds went from ~1 hour and 45 minutes to ~5 minutes.
Confounding variable to loading speeds: this is over a fiber-attached storage device. Connection is 10gbe full-duplex, remote source has an NVMe raid cache. Saturation of the network is the norm, but laintency is a factor. Regardless, small-scale local-storage testing also shows measurable improvements, so this should be a welcome addition.
2023-08-15 16:07:41 -04:00
79c0131158
Deploying to master from @ vladmandic/automatic@b1ea529c08 🚀
2023-08-15 12:25:08 +00:00
86ae8175e0
Seperate OpenVINO from IPEX
2023-08-15 15:22:54 +03:00
3d914688cc
update metadata
2023-08-15 05:50:15 +02:00
2caee04b9a
IPEX fix System Info
2023-08-16 19:45:13 +03:00
209f9a19c6
IPEX fixes
2023-08-16 18:56:50 +03:00
b1ea529c08
Cleanup
2023-08-15 00:50:46 +03:00
66394c8cd9
IPEX add OpenVINO as compile backend
2023-08-14 21:54:21 +03:00
fce48be440
update changelog
2023-08-14 07:58:42 +00:00
eeca263bd2
update stored metadata
2023-08-14 07:50:12 +00:00
88fff06c9e
downgrade warn to info
2023-08-13 10:58:02 +00:00
357b5dac60
add dpm++ 3m sde sampler
2023-08-13 10:19:33 +00:00
d4b5c487fe
fix img2img resize
2023-08-13 09:54:09 +00:00
c5c817f482
img2img batching
2023-08-13 09:19:39 +00:00
23efa8f0a5
fix init_image
2023-08-13 08:22:03 +00:00
7e81355741
Fix pdh.dll not found. (WSL+DirectML)
2023-08-13 15:40:15 +09:00
83d1ee77bd
revert long prompt
2023-08-13 00:32:03 +02:00
392f51cff3
add long prompts
2023-08-12 23:04:46 +02:00
69eaf4c664
fix batch
2023-08-12 08:32:19 +00:00
5d5f22e6e1
remove invalid downloaded files
2023-08-12 08:07:13 +00:00
fc8eca1c34
deallocate images on batch
2023-08-12 07:18:19 +00:00
605988f520
Removed annoyance
...
Took out the print statement 'no image data blocks found' as it happens frequently as embeddings rarely have image data.
2023-08-11 19:51:42 -04:00
a156751857
fix pipeline autodetect
2023-08-10 21:56:26 +00:00
f52249d5a8
fix prompt parser for sdxl and enable offloading
2023-08-10 21:20:56 +00:00
5bcd65d4c2
revert meta
2023-08-10 19:00:04 +00:00
0a7105d524
Fix SDXL LoRa offloading and SD 1.5 parsing
2023-08-10 16:02:45 +03:00
0b5dd46d79
Merge pull request #1976 from tcmaps/master
...
pip --no-cache-dir
2023-08-10 08:28:02 +02:00
a9726b3319
Send to meta when unloading
2023-08-10 01:23:39 +03:00
839bd27956
Pass device to Compel
2023-08-09 19:15:38 +03:00
16725ab38f
fix compel to full and add batch sizes
2023-08-09 10:28:44 +00:00
e2b0d981ac
quick taesd vae decode
2023-08-09 09:34:59 +00:00
a77fa4c12e
add job info
2023-08-09 08:42:26 +00:00
a5f95f4b38
civitai model search and download
2023-08-09 08:01:37 +00:00
196fa29ee0
Update ui_extra_networks.py
...
Added fix to thumbnail images that exceed 70kb to address slowdowns and wasted memory.
2023-08-08 11:43:38 -04:00
ff2d846667
IPEX Diffusers fix Model CPU Offload
2023-08-08 00:51:23 +03:00
6760fd1525
Add TAESD VAE option for base image outputs
2023-08-07 21:37:58 +03:00
0a3e821067
diffuser auto-pipeline and fix vae
2023-08-07 17:19:30 +00:00
9d17cf4c12
IPEX Diffusers fix can't allocate 4GB+ with SDP
2023-08-07 19:50:39 +03:00
5fd3c5ba01
update lora methods
2023-08-07 14:52:19 +02:00
9be641d431
Merge pull request #1952 from AI-Casanova/kohya
...
Enable side loading of multiple Kohya-style LoRA for SDXL
2023-08-07 14:42:30 +02:00
Vladimir Mandic
Midcoastal
vladmandic
Disty0
Seunghoon Lee
Aptronymist