Commit Graph

508 Commits

Author SHA1 Message Date
Jaret Burkett
bd8d7dc081 fixed various issues with llm attention masking. Added block training on the llm adapter. 2025-02-14 11:24:01 -07:00
Jaret Burkett
2be6926398 Added back syustem prompt for llm and remove those tokens from the embeddings 2025-02-14 07:23:37 -07:00
Jaret Burkett
87ac031859 Remove system prompt, shouldnt be necessary fo rhow it works. 2025-02-13 08:42:48 -07:00
Jaret Burkett
7679105d52 Added llm text encoder adapter 2025-02-13 08:28:32 -07:00
Jaret Burkett
2622de1e01 DFE tweaks. Adding support for more llms as text encoders 2025-02-13 04:31:49 -07:00
Jaret Burkett
8450aca10e Fixed missed merge conflice and locked diffusers version 2025-02-12 09:40:02 -07:00
Jaret Burkett
0b8a32def7 merged in lumina2 branch 2025-02-12 09:33:03 -07:00
Jaret Burkett
787bb37e76 Small fixed for DFE, polar guidance, and other things 2025-02-12 09:27:44 -07:00
Jaret Burkett
10aa7e9d5e Fixed some breaking changes with diffusers gradient checkpointing. 2025-02-10 09:35:31 -07:00
Jaret Burkett
ed1deb71c4 Added examples for training lumina2 2025-02-08 16:13:18 -07:00
Jaret Burkett
4de6a825fa Update lumina requirements 2025-02-08 15:16:35 -07:00
Jaret Burkett
9a7266275d Wokr on lumina2 2025-02-08 14:52:39 -07:00
Jaret Burkett
d138f07365 Imitial lumina3 support 2025-02-08 10:59:53 -07:00
Jaret Burkett
c6d8eedb94 Added ability to use consistent noise for each image in a dataset by hashing the path and using that as a seed. 2025-02-08 07:13:48 -07:00
Jaret Burkett
af5e760be1 Merge pull request #249 from ostris/accelerate-multi-gpu
Multi gpu support. Other goodies
2025-02-08 07:11:10 -07:00
Jaret Burkett
ff3d54bb5b Make the mean of the mask multiplier be 1.0 for a more balanced loss. 2025-02-06 10:57:06 +00:00
Jaret Burkett
0e75724b4d Lock version of diffusers 2025-02-05 18:14:59 +00:00
Jaret Burkett
376bb1bf6f Lock torch version due to breaking changes 2025-02-04 23:00:57 +00:00
Jaret Burkett
216ab164ce Experimental features and bug fixes 2025-02-04 13:36:34 -07:00
Jaret Burkett
e6180d1e1d Bug fixes 2025-01-31 13:23:01 -07:00
Jaret Burkett
15a57bc89f Add new version of DFE. Kitchen sink 2025-01-31 11:42:27 -07:00
Jaret Burkett
e5355bf8d5 Added train catch to the blank network 2025-01-30 16:15:45 +00:00
Jaret Burkett
34a1c6947a Added flux_shift as timestep type 2025-01-27 07:35:00 -07:00
Jaret Burkett
2141c6e06c Merge remote-tracking branch 'origin/main' into accelerate-multi-gpu 2025-01-26 11:19:34 -07:00
Jaret Burkett
1188cf1e8a Adjust flux sample sampler to handle some new breaking changes in diffusers. 2025-01-26 18:09:21 +00:00
Jaret Burkett
5e663746b8 Working multi gpu training. Still need a lot of tweaks and testing. 2025-01-25 16:46:20 -07:00
Jaret Burkett
441474e81f Added a flag to lora extraction script to do a full transformer extraction. 2025-01-24 09:34:13 -07:00
Jaret Burkett
a6a690f796 Update full fine tune example to only train transformer blocks. 2025-01-24 09:28:34 -07:00
Jaret Burkett
6191f19e55 Added script to convert diffusers model to ComfyUI variant 2025-01-23 21:30:23 -07:00
Jaret Burkett
bbfba0c188 Added v2 of dfp 2025-01-22 16:32:13 -07:00
Jaret Burkett
e1549ad54d Update dfe model arch 2025-01-22 10:37:23 -07:00
Jaret Burkett
04abe57c76 Added weighing to DFE 2025-01-22 08:50:57 -07:00
Jaret Burkett
89dd041b97 Added ability to pair samples with a closer noise with optimal_noise_pairing_samples 2025-01-21 18:30:10 -07:00
Jaret Burkett
29122b1a54 Added code to handle diffusion feature extraction loss 2025-01-21 14:21:34 -07:00
Jaret Burkett
6a8e3d8610 Added a config file for full finetuning flex. Added a lora extraction script for flex 2025-01-20 10:09:01 -07:00
Jaret Burkett
4c8a9e1b88 Added example config to train Flex 2025-01-18 18:03:20 -07:00
Jaret Burkett
fadb2f3a76 Allow quantizing the te independently on flux. added lognorm_blend timestep schedule 2025-01-18 18:02:31 -07:00
Jaret Burkett
4723f23c0d Added ability to split up flux across gpus (experimental). Changed the way timestep scheduling works to prep for more specific schedules. 2024-12-31 07:06:55 -07:00
Jaret Burkett
8ef07a9c36 Added training for an experimental decoratgor embedding. Allow for turning off guidance embedding on flux (for unreleased model). Various bug fixes and modifications 2024-12-15 08:59:27 -07:00
Jaret Burkett
92ce93140e Adjustments to defaults for automagic 2024-11-29 10:28:06 -07:00
Jaret Burkett
f213996aa5 Fixed saving and displaying for automagic 2024-11-29 08:00:22 -07:00
Jaret Burkett
cbe31eaf0a Initial work on a auto adjusting optimizer 2024-11-29 04:48:58 -07:00
Jaret Burkett
67c2e44edb Added support for training flux redux adapters 2024-11-21 20:01:52 -07:00
Jaret Burkett
96d418bb95 Added support for full finetuning flux with randomized param activation. Examples coming soon 2024-11-21 13:05:32 -07:00
Jaret Burkett
894374b2e9 Various bug fixes and optimizations for quantized training. Added untested custom adam8bit optimizer. Did some work on LoRM (dont use) 2024-11-20 09:16:55 -07:00
Jaret Burkett
6509ba4484 Fix seed generation to make it deterministic so it is consistant from gpu to gpu 2024-11-15 12:11:13 -07:00
Jaret Burkett
025ee3dd3d Added ability for adafactor to fully fine tune quantized model. 2024-10-30 16:38:07 -06:00
Jaret Burkett
58f9d01c2b Added adafactor implementation that handles stochastic rounding of update and accumulation 2024-10-30 05:25:57 -06:00
Jaret Burkett
e72b59a8e9 Added experimental 8bit version of prodigy with stochastic rounding and stochastic gradient accumulation. Still testing. 2024-10-29 14:28:28 -06:00
Jaret Burkett
4aa19b5c1d Only quantize flux T5 is also quantizing model. Load TE from original name and path if fine tuning. 2024-10-29 14:25:31 -06:00