Commit Graph

130 Commits

Author SHA1 Message Date
Falko Victor Habel 86664b10a6 improved vram usage? 2025-02-23 22:26:48 +01:00
Falko Victor Habel 7603ce8851 massivly downsized data 2025-02-23 20:09:28 +01:00
Falko Victor Habel ad27ea2fa2 fixed to load kwargs 2025-02-23 20:04:44 +01:00
Falko Victor Habel 74c4adbff0 debug print 2025-02-23 20:02:31 +01:00
Falko Victor Habel 86bb74835d updated Upsampler config 2025-02-23 19:59:18 +01:00
Falko Victor Habel 048a8d9861 added extra config 2025-02-23 19:56:17 +01:00
Falko Victor Habel b740819757 fixed paths 2025-02-23 19:27:34 +01:00
Falko Victor Habel e52a2a4b81 fixed spelling 2025-02-23 19:26:55 +01:00
Falko Victor Habel 2883e67e1c temporary commit for lowercase fix 2025-02-23 19:26:36 +01:00
Falko Victor Habel 736886021c updated finetuning script to work with Upsamler and added Early Stopping 2025-02-22 17:53:06 +01:00
Falko Victor Habel 2aba93caea updated upsampler 2025-02-22 17:52:37 +01:00
Falko Victor Habel 933238a530 added extra psampling layer [not working] 2025-02-22 17:24:02 +01:00
Falko Victor Habel 90bcdd346a even more downsized 2025-02-22 10:45:30 +01:00
Falko Victor Habel f576a0d789 fixed autocast 2025-02-22 10:43:10 +01:00
Falko Victor Habel 81ceddff3b addded device gpu and fixed autocast 2025-02-21 22:01:54 +01:00
Falko Victor Habel 742a3e6f70 fiexd script 2025-02-21 21:59:16 +01:00
Falko Victor Habel 4af12873f2 cpu training 2025-02-21 21:54:33 +01:00
Falko Victor Habel 992bc0eb7f downsized two 8k samples 2025-02-21 21:51:48 +01:00
Falko Victor Habel 0166d8feeb single batch 2025-02-21 21:49:16 +01:00
Falko Victor Habel b06dad23af downsized samples to max of 10k 2025-02-21 21:47:06 +01:00
Falko Victor Habel 6a31e8fa7f downsized to 2 2025-02-21 21:43:01 +01:00
Falko Victor Habel d66b167b85 downsized abtchsize to 8 2025-02-21 21:17:40 +01:00
Falko Victor Habel fca74fb8d2 new loading 2025-02-21 21:12:14 +01:00
Falko Victor Habel 2583d2f01f bugfix 2025-02-21 21:05:48 +01:00
Falko Victor Habel 2f12fcb863 updated loading 2025-02-21 21:03:37 +01:00
Falko Victor Habel 6a9b4afd91 use correct model loading 2025-02-21 20:58:35 +01:00
Falko Victor Habel eacef6af65 added tqdm 2025-02-21 20:57:11 +01:00
Falko Victor Habel 36f9b30666 new saving 2025-02-21 20:55:46 +01:00
Falko Victor Habel 0c0372794e new alternative version 2025-02-21 20:54:30 +01:00
Falko Victor Habel 40f1a65006 fix 2025-02-17 16:31:12 +01:00
Falko Victor Habel bba5e2179d setted up correct amp 2025-02-17 16:26:54 +01:00
Falko Victor Habel e28fd3d330 added paths 2025-02-17 16:24:14 +01:00
Falko Victor Habel 704ad6106d updated inference class and test image 2025-02-17 16:22:47 +01:00
Falko Victor Habel 62825e9731 increased image size 2025-02-15 13:29:50 +01:00
Falko Victor Habel 4c84086932 updated script 2025-02-15 13:27:30 +01:00
Falko Victor Habel 619e17c32c going with batchsize 1 2025-02-15 13:17:17 +01:00
Falko Victor Habel e69d0e90ec fixed checkpoints tensors 2025-02-15 13:08:09 +01:00
Falko Victor Habel 75be3291d3 added custom upscaler model 2025-02-14 22:39:20 +01:00
Falko Victor Habel 1234fc5beb added device type to autocast 2025-02-14 22:27:19 +01:00
Falko Victor Habel ca29ee748b fixed rsults 2025-02-14 22:23:38 +01:00
Falko Victor Habel 9645e1da23 fixed checkpoint 2025-02-14 22:14:09 +01:00
Falko Victor Habel 5a6680178a try checkpoints 2025-02-14 22:10:07 +01:00
Falko Victor Habel e7e7e96001 max gpu usage added 2025-02-14 22:03:04 +01:00
Falko Victor Habel ca44dd8a77 added torch.amp support 2025-02-14 21:48:52 +01:00
Falko Victor Habel b4dd550f8d downsized params 2025-02-14 21:33:47 +01:00
Falko Victor Habel 19e5b72724 simplified code 2025-02-12 06:59:18 +01:00
Falko Victor Habel 8fafbebe45 fixed loading 2025-02-08 16:55:54 +01:00
Falko Victor Habel 20dd9f68ed added tqdm to improve code 2025-02-08 12:52:48 +01:00
Falko Victor Habel a48bad5a49 limit dataset 2025-02-07 15:26:57 +01:00
Falko Victor Habel 39efdff09e updated datasets 2025-02-06 20:54:05 +01:00