Mirrors
/
EveryDream2trainer
mirror of https://github.com/victorchall/EveryDream2trainer.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
820 Commits 6 Branches 0 Tags 5.4 MiB
Commit Graph

820 Commits

Author SHA1 Message Date
Victor Hall 42ad271261 bug in caption cog if starts_with no there 2024-03-02 01:58:13 -05:00
Victor Hall d098223052 merge conflict 2024-03-02 01:20:03 -05:00
Victor Hall be2cec7d3b
Update Train_Colab.ipynb 2024-02-23 11:33:09 -05:00
BlueBug e1e070ca83 Cough... fixed typo 
me not trying to be a grammar villian but a helpful noob
 2024-02-06 18:11:20 +09:00
BlueBug eee130d7dd Fixs error when translated caption contains an unusual byte 
fixs this
File "D:\EveryDream2trainer\caption_cog.py", line 143, in main
    f.write(caption)
UnicodeEncodeError: 'cp949' codec can't encode character '\xe9' in position 333: illegal multibyte sequence
 2024-02-06 17:49:21 +09:00
Victor Hall 5622b55128 fix up cog caption script 2024-02-03 22:01:14 -05:00
Victor Hall 1a8495a706 cleanup on cog caption script 2024-02-03 19:25:20 -05:00
Victor Hall cbc9a2d337
add cog-vlm to citations 2024-02-03 14:52:16 -05:00
Victor Hall f885be56a1 add cog caption script 2024-01-24 22:21:51 -05:00
Victor Hall 18198715b7
Merge pull request #246 from damian0815/patch-2 
prevent OOM with disabled unet when gradient checkpointing is enabled
 2024-01-15 20:11:52 -05:00
Damian Stewart 9fc6ae7a09
prevent OOM with disabled unet when gradient checkpointing is enabled 
unet needs to be in train() mode for gradient checkpointing to work
 2024-01-16 10:23:52 +13:00
Victor Hall e08d5ded98
Merge pull request #245 from Enixile/main 
BLIP2 and GIT updates
 2024-01-10 13:47:08 -05:00
Enixile 817f4169ce BLIP2 was loading into system RAM rather than VRAM for me. I found that adding argument "device_map=device" forces it to load the entire model onto the specified device. I tested it for cuda and cpu successfully. 
GIT was failing because its processor returns a BatchEncoding and its method to() only accepts one device parameter, so I removed the second dtype parameter and it worked. I don't see how BLIP2 is different, but it works with both parameters, so I left it to avoid breaking any functionality.
 2024-01-06 15:16:59 -05:00
Victor Hall e8f7b6c8b4 fix oopsie 2023-12-22 13:54:20 -05:00
Victor Hall e89ec053c7 fix issue with plugin interaction with validation 2023-12-22 13:52:26 -05:00
Victor Hall 9141710d6e add ps activate script 2023-12-21 14:02:36 -05:00
Victor Hall c7d3064029 error in dataloader plugin 2023-12-21 10:50:11 -05:00
Victor Hall 7a81182220
Update PLUGINS.md 
update plugin docs
 2023-12-20 16:46:44 -05:00
Victor Hall 1119c2130f
Update PLUGINS.md 2023-12-20 15:39:57 -05:00
Victor Hall 83623929ad
Update PLUGINS.md 
add docs for dataloader plugin hooks
 2023-12-20 15:29:33 -05:00
Victor Hall d9d2a43804 fix order of running pil image dataloader plugin 2023-12-20 15:24:09 -05:00
Victor Hall dfcc9e7f41 add dataloader plugin hooks for caption and pil image 2023-12-20 14:55:50 -05:00
Victor Hall 1236329677 fix reshaping for picking apps 2023-11-29 22:17:52 -05:00
Victor Hall a05ffca82e add positions id back for clip-l 2023-11-29 20:25:38 -05:00
Victor Hall 49e382c82b script to backdate autoencoderkl keys due to other broken software 2023-11-29 19:02:50 -05:00
Victor Hall 3935a49730 removing open flamingo from standard installs and documenting, OF is broken in torch2.1 unfortunately 2023-11-26 12:51:58 -05:00
Victor Hall 670354de94 removing open flamingo from requirements.txt 2023-11-26 12:33:04 -05:00
Victor Hall af49f3ce08
Merge pull request #243 from victorchall/attn 
Add SDP attention and update windows to torch 2.1
 2023-11-25 19:09:36 -05:00
Victor Hall 8756eb9863 Merge branch 'attn' of https://github.com/victorchall/EveryDream2trainer into attn 2023-11-25 18:47:53 -05:00
Victor Hall ec98a79b1e add tokenizer tester script 2023-11-25 17:21:34 -05:00
Victor Hall e705db29eb doc for attn_type 2023-11-25 13:43:26 -05:00
Victor Hall a284c52dee update windows to torch 2.1 and add attention type option 2023-11-24 16:13:01 -05:00
Victor Hall 772d6bc439
Update ADVANCED_TWEAKING.md 2023-11-21 14:07:21 -05:00
Victor Hall 051ecf57ad
Merge pull request #242 from a-l-e-x-d-s-9/main 
New optimizer and pyramid noise
 2023-11-21 14:02:58 -05:00
Victor Hall af74c7f5e1 jupyter notebook fix to upload ldm safetensors instead of ckpt 2023-11-18 13:33:22 -05:00
alexds9 32ee7cb7f0 Using -1.0 instead of -1. 2023-11-17 21:55:12 +02:00
Alex dcf2969640
Merge branch 'victorchall:main' into main 2023-11-17 20:52:34 +02:00
alexds9 fc6a4a7c27 LR removed from AdaCoor. 2023-11-17 20:52:15 +02:00
Victor Hall 31e1619a13 clean up citations 2023-11-17 13:39:12 -05:00
Victor Hall e3e542c121 add some citations and references links 2023-11-17 12:56:53 -05:00
Victor Hall d2dac6fd06 update requirements.txt to have newer xformers 2023-11-17 12:31:25 -05:00
Victor Hall 6727b6d61f fix log_writer bug and move logs into specific project log folder 2023-11-17 12:30:57 -05:00
alexds9 5dc9f18061 1. Added AdaCoor optimizer. 
2. Added pyramid noise.
3. Fixed problem with log_writer missing from EveryDreamOptimizer.
 2023-11-17 16:08:43 +02:00
Victor Hall bf3c022489 update docker tags 2023-11-16 12:04:39 -05:00
Victor Hall 840493037e gg bug from pr that thought it was simplifying code but broke it 2023-11-15 16:19:08 -05:00
Victor Hall 097d864ef5 fixing bugs 2023-11-15 15:57:09 -05:00
Victor Hall 20a9b3254f bug in optimizer decay-warmup defaulting 2023-11-15 15:46:28 -05:00
Victor Hall e95a8861e9 workaround for cuda errors on some rented machines 2023-11-15 14:45:43 -05:00
Victor Hall a95d2de150 fix ckpt to diffusers conversion script 2023-11-15 12:55:22 -05:00
Victor Hall b63ae0dafd
Merge pull request #238 from reijerh/patch-1 
Quick edits to improve PLUGINS.md
 2023-11-13 13:41:10 -05:00