finetune_class #1

Fabel · 2025-02-26T12:13:04Z

Fabel commented

2025-02-26 12:13:04 +00:00

No description provided.

Fabel added 96 commits 2025-02-26 12:13:05 +00:00

d85faadcc1 first finetune script

71da7ed2f1 removed old scripts

914d002602 added new aiiun script with first draft for pip project

4a60045320 first finetune try

2121316e3b finetune improvement

73d52f733c updated config for updated model

1f33f22bea convert the bytes first

be5bb53620 working transform

4037a07764 updated process

9b7a182782 debug mode

34c547fb23 next try

0484ae01b1 added tqdm and removed doubles

8c49cc7e01 fixed tqdm in finetuning

9ec563e86d updated Finetuner to correctly log losses

13fb2b76c1 updated finetuning

6b35463027 upgaeed path for base model

0b04ebab7a corected path again

804882d7a2 corrected dataset paths

50fa103579 printing

325faef9d5 push

9187ebe012 debug print

be74658ceb added decoeder model

bb22d0a6da added decoder models fits

de6a67cb4e updated finetuning script

e298ac4c45 manually loading the correct modle again

1a66169b37 updated keys

72a5959bc1 updated batchsize and fixed key names

bac4a9010a updated dataset loading

39efdff09e updated datasets

a48bad5a49 limit dataset

20dd9f68ed added tqdm to improve code

8fafbebe45 fixed loading

19e5b72724 simplified code

b4dd550f8d downsized params

ca44dd8a77 added torch.amp support

e7e7e96001 max gpu usage added

5a6680178a try checkpoints

9645e1da23 fixed checkpoint

ca29ee748b fixed rsults

1234fc5beb added device type to autocast

75be3291d3 added custom upscaler model

e69d0e90ec fixed checkpoints tensors

619e17c32c going with batchsize 1

4c84086932 updated script

62825e9731 increased image size

704ad6106d updated inference class and test image

e28fd3d330 added paths

bba5e2179d setted up correct amp

40f1a65006 fix

0c0372794e new alternative version

36f9b30666 new saving

eacef6af65 added tqdm

6a9b4afd91 use correct model loading

2f12fcb863 updated loading

2583d2f01f bugfix

fca74fb8d2 new loading

d66b167b85 downsized abtchsize to 8

6a31e8fa7f downsized to 2

b06dad23af downsized samples to max of 10k

0166d8feeb single batch

992bc0eb7f downsized two 8k samples

4af12873f2 cpu training

742a3e6f70 fiexd script

81ceddff3b addded device gpu and fixed autocast

f576a0d789 fixed autocast

90bcdd346a even more downsized

933238a530 added extra psampling layer [not working]

2aba93caea updated upsampler

736886021c updated finetuning script to work with Upsamler and added Early Stopping

2883e67e1c temporary commit for lowercase fix

e52a2a4b81 fixed spelling

b740819757 fixed paths

048a8d9861 added extra config

86bb74835d updated Upsampler config

74c4adbff0 debug print

ad27ea2fa2 fixed to load kwargs

7603ce8851 massivly downsized data

86664b10a6 improved vram usage?

c88961aee7 convert back to images

a51300c77c base model fix

fde8bdcb6f test

e114023cbc go to cpu

8935fa5e13 doubled config

33a3626b74 fixed config

b89670ad6c loaded with bf16

0fa14c4a6f added gpu support

443b9f5589 use cpu

2360e23cc7 downsized imageset

64fb7a199d antialisaing not available

ecb2694415 update color channels

153ead15b7 downsized image amount

79e0aeb269 downsized batchsize increased images

0611a26564 updated project info

09f196294c new upsampler scripts

fcebc103b8 fixed paths

6aad1a38f5 increased trainingset amount

Fabel merged commit 930f3a4885 into develop

2025-02-26 12:13:09 +00:00

Fabel referenced this issue from a commit

2025-02-26 12:13:10 +00:00

Merge pull request 'finetune_class' (#1) from finetune_class into develop

Fabel deleted branch finetune_class

2025-02-26 12:13:12 +00:00

Sign in to join this conversation.

No reviewers

No Label

No Milestone

No project

No Assignees

1 Participants

Notifications

Due Date

The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: Machine-Learning/aiuNN#1