Commit Graph

Select branches

Hide Pull Requests

main

old

sd3

ti_cross_attn_loss

Mono Color

• b379a28715 update todos mayukhdeb 2024-06-22 01:38:12 -07:00
• 54ff8c4977 sd3 concept inference mayukhdeb 2024-06-22 01:25:16 -07:00
• 64f28c8589 save TI embeds and lora adapters mayukhdeb 2024-06-22 01:24:30 -07:00
• b12ce26fc9 update command mayukhdeb 2024-06-20 03:52:30 -07:00
• b3da65dd39 ignore wandb stuff mayukhdeb 2024-06-20 03:46:26 -07:00
• f64da5d4f1 update todo mayukhdeb 2024-06-20 03:44:42 -07:00
• 5cc09092a4 tweak param mayukhdeb 2024-06-20 03:43:44 -07:00
• 3d3d0ee4cb smash more todos mayukhdeb 2024-06-20 03:42:32 -07:00
• 4b1efce0e3 compute loss and update weights mayukhdeb 2024-06-20 03:28:49 -07:00
• d115c52274 do just forward passes mayukhdeb 2024-06-20 03:20:41 -07:00
• 22127c9917 typo mayukhdeb 2024-06-20 01:41:28 -07:00
• 4844845d5c more progress mayukhdeb 2024-06-20 01:40:59 -07:00
• 3b15250bbc init train dataloader + update todos for training mayukhdeb 2024-06-20 00:54:41 -07:00
• fd448e5437 accomodate T5EncoderModel mayukhdeb 2024-06-20 00:54:21 -07:00
• a41cce1486 small cleanup mayukhdeb 2024-06-20 00:27:51 -07:00
• 12b5960c36 progress bar for latent caching mayukhdeb 2024-06-20 00:27:21 -07:00
• 2c2bfcda13 init PreprocessedDataset mayukhdeb 2024-06-20 00:27:01 -07:00
• cacd6f2201 count trainable params from model mayukhdeb 2024-06-19 23:58:26 -07:00
• 67024b4de6 full or lora finetuning of sd3 transformer mayukhdeb 2024-06-19 23:58:16 -07:00
• 6d0d96ec79 more progress on todos mayukhdeb 2024-06-19 23:37:31 -07:00
• fb2449a60d init textual inversion token embeds mayukhdeb 2024-06-17 07:31:54 -07:00
• 0496ccdbef handle sd3 t5 text encoder mayukhdeb 2024-06-17 07:31:33 -07:00
• 044aed9f03 sd3 train script wip mayukhdeb 2024-06-17 06:42:16 -07:00
• bdda796c56 ignore notebook checkpoint mayukhdeb 2024-06-17 05:21:54 -07:00
• b02221b23c cleanup before sd3 integration xander 2024-06-13 14:24:32 +02:00
• d8b3175b57 cleanup before sd3 integration xander 2024-06-13 13:57:20 +02:00
• a80c934679 add name aiXander 2024-06-05 05:47:26 -07:00
• 813e95592d initial commit of ComfyUI node aiXander 2024-06-05 03:30:18 -07:00
• 41ddfbcca6 add resource monitoring xander 2024-05-29 23:33:42 +02:00
• 8cd935251d add strip urls xander 2024-05-28 23:11:41 +02:00
• 0ed83dc656 update readme xander 2024-05-28 17:33:13 +02:00
• 3ebbcf726b Update README.md Xander Steenbrugge 2024-05-28 17:28:31 +02:00
• 3883adb697 Update README.md Xander Steenbrugge 2024-05-28 17:28:15 +02:00
• 2033b13370 update readme xander 2024-05-28 17:26:46 +02:00
• 3dabbfd55c add description for modes xander 2024-05-28 17:11:11 +02:00
• 0d8f5ca929 update readme xander 2024-05-28 17:09:30 +02:00
• 128adb99ab add imgs xander 2024-05-28 17:07:55 +02:00
• 46ce69a29a update readme xander 2024-05-28 17:04:17 +02:00
• 135959ba0e update readme xander 2024-05-28 16:56:56 +02:00
• c845c76b15 update readme xander 2024-05-28 16:55:49 +02:00
• f6f50e5004 merge xander 2024-05-28 16:51:43 +02:00
• e9b5ca0dcc small tweaks xander 2024-05-28 16:51:02 +02:00
• eb144d5212 increase starting lr for sd15 aiXander 2024-05-16 16:46:08 -07:00
• 3f7ba82a25 tiny fixes aiXander 2024-05-16 16:02:46 -07:00
• 6e32be4dfd bugfix aiXander 2024-05-16 14:31:57 -07:00
• 3e17bed208 update workflows aiXander 2024-05-16 14:03:14 -07:00
• 9102e8e5e5 render with active pipe when not in debug mode aiXander 2024-05-16 09:28:22 -07:00
• 03c6175007 fallback to blip when there are many imgs to caption aiXander 2024-05-16 09:04:20 -07:00
• 13ee2ad6b5 plot categorical labels xander 2024-05-16 16:38:52 +02:00
• e665169ce8 use gpt4-o aiXander 2024-05-16 07:37:02 -07:00
• de99482825 fix predict cog bugs aiXander 2024-05-15 17:43:17 -07:00
• 0d0ad2a4d0 update base model aiXander 2024-05-15 16:54:49 -07:00
• 33b4ac7cc2 update default settings aiXander 2024-05-15 16:42:55 -07:00
• 3f7298633d update cog run settings aiXander 2024-05-15 14:59:48 -07:00
• fec9cbfadd fix cog predict aiXander 2024-05-14 15:11:07 -07:00
• 06b3b7b459 print debug state aiXander 2024-05-14 07:07:13 -07:00
• 70a645ac79 fix bugs aiXander 2024-05-14 07:04:25 -07:00
• fa76a90c39 add workflows xander 2024-05-14 15:25:35 +02:00
• 8b3f9a930f rename workflows xander 2024-05-14 15:24:33 +02:00
• cfde9e9180 tiny param tweak xander 2024-05-14 15:23:26 +02:00
• 4389ae75c8 force quit training after max steps xander 2024-05-14 15:21:48 +02:00
• d66863b4cb more cleanup xander 2024-05-14 15:12:49 +02:00
• 110ec0df30 fix lora_scale inference validation samples xander 2024-05-14 14:34:57 +02:00
• b51727219e cleanup and try to fix last inference step xander 2024-05-14 11:42:57 +02:00
• b8fe89e983 fix typo xander 2024-05-02 17:11:54 +02:00
• e47fcf429e update predict.py aiXander 2024-05-02 07:59:13 -07:00
• fc63399317 updates xander 2024-05-02 16:53:49 +02:00
• 0bd26033cd add conv2 to lora layers xander 2024-04-29 20:12:08 +02:00
• 7d78a3d76d updates xander 2024-04-29 19:37:45 +02:00
• 45e6f50c24 small updates xander 2024-04-29 12:10:17 +02:00
• 39ee9c9630 set defaults xander 2024-04-27 16:47:35 +02:00
• 24fcc2e791 update defaults xander 2024-04-26 15:59:18 +02:00
• 8da2f58c20 everything is working xander 2024-04-26 14:52:32 +02:00
• 7d469d843c cleanup and make all this work also for sd15 xander 2024-04-26 14:38:25 +02:00
• 2e5912f629 improve token regularization xander 2024-04-26 13:32:01 +02:00
• 37bcfe61c9 hyperparam tweaks xander 2024-04-26 08:45:29 +02:00
• af6c4676d5 bugfixes xander 2024-04-25 21:41:57 +02:00
• 9ae9bc19d2 mask improvements xander 2024-04-25 21:21:14 +02:00
• 7218c7e8c2 more tok reg updates xander 2024-04-24 17:51:42 +02:00
• c60f3646fa update tok_reg loss xander 2024-04-24 16:43:22 +02:00
• e1b05b1b99 update loss plots, reduce tok_reg value xander 2024-04-24 14:42:05 +02:00
• 0a72748da6 update preprocess to work better for 1 img xander 2024-04-23 16:21:49 +02:00
• f919e3e7a1 setup experiment run xander 2024-04-23 04:39:35 +02:00
• d4923ab6d2 tiny typo fix aiXander 2024-04-22 19:39:16 -07:00
• d613eff061 more tweaks and cleanup aiXander 2024-04-22 19:24:28 -07:00
• b48419247a refactor token regularization aiXander 2024-04-22 19:08:36 -07:00
• 4d9f8c2e67 more tweaks aiXander 2024-04-22 12:22:20 -07:00
• 34e4bc227c ready for pushing new concept trainer aiXander 2024-04-22 12:00:40 -07:00
• c85c45b688 add a few more params to cog input aiXander 2024-04-22 11:41:37 -07:00
• d4dbb52f9b add adapter_config.json save aiXander 2024-04-22 11:32:46 -07:00
• 784bfbf14f prep for cog push aiXander 2024-04-22 11:27:04 -07:00
• 327339ebba prep for cog push aiXander 2024-04-22 08:41:00 -07:00
• f2d70b64da add dockerignore aiXander 2024-04-22 05:48:19 -07:00
• 41b7895bff add unet_prodigy_growth_factor xander 2024-04-22 14:36:04 +02:00
• 5b4020d448 save sd15 models as they're supposed to be saved mayukhdeb 2024-04-20 06:21:15 -07:00
• 2633c06a87 update gpt4 model aiXander 2024-04-19 07:36:51 -07:00
• a33a33295e never load from disk during training mayukhdeb 2024-04-19 05:33:07 -07:00
• 3e44dff1da remove pipe before loading from disk aiXander 2024-04-19 05:19:49 -07:00
• b2080da7c8 cleanup checkpoint aiXander 2024-04-19 05:08:47 -07:00
• 13733a4f19 Merge branch 'main' of https://github.com/edenartlab/trainer into main aiXander 2024-04-19 04:58:09 -07:00