Version 0.8.6 #1212
Version 0.8.6 #1212
Commits on Feb 3, 2024
Commits on Feb 5, 2024
Commits on Feb 9, 2024
-
the reason not working grad accum steps found. it was becasue of my a…
…ccelerate settings
Commits on Feb 22, 2024
Commits on Feb 27, 2024
-
-
-
Merge pull request #1101 from BootsofLagrangian/deepspeed
support deepspeed
-
-
Commits on Mar 3, 2024
Commits on Mar 11, 2024
Commits on Mar 20, 2024
-
-
-
Merge pull request #1177 from KohakuBlueleaf/random-strength-noise
Random strength for Noise Offset and input perturbation noise
-
English Translation of config_README-ja.md (#1175)
* Add files via upload Creating template to work on. * Update config_README-en.md Total Conversion from Japanese to English. * Update config_README-en.md * Update config_README-en.md * Update config_README-en.md
-
-
Merge pull request #1167 from Horizon1704/patch-1
Add "encoding='utf-8'" for --config_file
-
-
Merge pull request #1168 from gesen2egee/save_state_on_train_end
Save state on train end
-
-
-
-
Fix most of ZeRO stage uses optimizer partitioning
- we have to prepare optimizer and ds_model at the same time. - pull/1139#issuecomment-1986790007 Signed-off-by: BootsofLagrangian <[email protected]>
-
Merge pull request #1200 from BootsofLagrangian/deep-speed
Fix sdxl_train.py in deepspeed branch
-
Commits on Mar 24, 2024
-
-
-
Merge pull request #1205 from feffy380/patch-1
register reg images with correct subset
-
-
[Experimental] Add cache mechanism for dataset groups to avoid long w…
…aiting time for initilization (#1178) * support meta cached dataset * add cache meta scripts * random ip_noise_gamma strength * random noise_offset strength * use correct settings for parser * cache path/caption/size only * revert mess up commit * revert mess up commit * Update requirements.txt * Add arguments for meta cache. * remove pickle implementation * Return sizes when enable cache --------- Co-authored-by: Kohya S <[email protected]>
-
-
-
-
-
-
Commits on Mar 26, 2024
-
-
-
Merge pull request #1206 from kohya-ss/dataset-cache
Add metadata caching for DreamBooth dataset
-
Commits on Mar 27, 2024
-
-
Merge pull request #1213 from Disty0/dev
Add OpenVINO and ROCm ONNX Runtime for WD14
-
Commits on Mar 28, 2024
Commits on Mar 29, 2024
Commits on Mar 30, 2024
-
-
-
Merge pull request #1216 from Disty0/dev
Rating support for WD Tagger
-
-
-
Commits on Mar 31, 2024
Commits on Apr 3, 2024
Commits on Apr 4, 2024
Commits on Apr 7, 2024
-
Merge pull request #1240 from kohya-ss/verify-command-line-args
verify command line args if wandb is enabled
-
-
Add option to use Scheduled Huber Loss in all training pipelines to i…
…mprove resilience to data corruption (#1228) * add huber loss and huber_c compute to train_util * add reduction modes * add huber_c retrieval from timestep getter * move get timesteps and huber to own function * add conditional loss to all training scripts * add cond loss to train network * add (scheduled) huber_loss to args * fixup twice timesteps getting * PHL-schedule should depend on noise scheduler's num timesteps * *2 multiplier to huber loss cause of 1/2 a^2 conv. The Taylor expansion of sqrt near zero gives 1/2 a^2, which differs from a^2 of the standard MSE loss. This change scales them better against one another * add option for smooth l1 (huber / delta) * unify huber scheduling * add snr huber scheduler --------- Co-authored-by: Kohya S <[email protected]>
-
-
