VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-09 06:45:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Kohya S.
2024-08-29 21:25:30 +09:00
committed by GitHub
parent a0cfb0894c
commit daa6ad5165
1 changed files with 6 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
README.md
View File

@@ -77,9 +77,9 @@ There are many unknown points in FLUX.1 training, so some settings can be specif
The existing `--loss_type` option may be useful for FLUX.1 training. The default is `l2`. The existing `--loss_type` option may be useful for FLUX.1 training. The default is `l2`.
~~In our experiments, `--timestep_sampling sigma --model_prediction_type raw --discrete_flow_shift 1.0` with `--loss_type l2` seems to work better than the default (SD3) settings. The multiplier of LoRA should be adjusted. ~~ ~~In our experiments, `--timestep_sampling sigma --model_prediction_type raw --discrete_flow_shift 1.0` with `--loss_type l2` seems to work better than the default (SD3) settings. The multiplier of LoRA should be adjusted.~~
In our experiments, `--timestep_sampling shift --discrete_flow_shift 3.1582 --model_prediction_type raw --guidance_scale 1.0` with `--loss_type l2` seems to work better than other settings. In our experiments, `--timestep_sampling shift --discrete_flow_shift 3.1582 --model_prediction_type raw --guidance_scale 1.0` (with the default `l2` loss_type) seems to work better.
The settings in [AI Toolkit by Ostris](https://github.com/ostris/ai-toolkit) seems to be equivalent to `--timestep_sampling sigmoid --model_prediction_type raw --guidance_scale 1.0` (with the default `l2` loss_type). The settings in [AI Toolkit by Ostris](https://github.com/ostris/ai-toolkit) seems to be equivalent to `--timestep_sampling sigmoid --model_prediction_type raw --guidance_scale 1.0` (with the default `l2` loss_type).
@@ -92,10 +92,13 @@ Other options are described below.
`--timestep_sampling` and `--sigmoid_scale`, `--discrete_flow_shift` adjust the distribution of timesteps. The distribution is shown in the figures below. `--timestep_sampling` and `--sigmoid_scale`, `--discrete_flow_shift` adjust the distribution of timesteps. The distribution is shown in the figures below.
The effect of `--discrete_flow_shift` with `--timestep_sampling shift` (when `--sigmoid_scale` is not specified, the default is 1.0): The effect of `--discrete_flow_shift` with `--timestep_sampling shift` (when `--sigmoid_scale` is not specified, the default is 1.0):
![Figure_2](https://github.com/user-attachments/assets/d9de42f9-f17d-40da-b88d-d964402569c6)
The difference between `--timestep_sampling uniform` and `--timestep_sampling sigma`: The difference between `--timestep_sampling sigmoid` and `--timestep_sampling uniform` (when `--timestep_sampling sigmoid` or `uniform` is specified, `--discrete_flow_shift` is ignored):
![Figure_3](https://github.com/user-attachments/assets/27029009-1f5d-4dc0-bb24-13d02ac4fdad)
The effect of `--timestep_sampling sigmoid` and `--sigmoid_scale` (when `--timestep_sampling sigmoid` is specified, `--discrete_flow_shift` is ignored): The effect of `--timestep_sampling sigmoid` and `--sigmoid_scale` (when `--timestep_sampling sigmoid` is specified, `--discrete_flow_shift` is ignored):
![Figure_4](https://github.com/user-attachments/assets/08a2267c-e47e-48b7-826e-f9a080787cdc)
#### Key Features for FLUX.1 LoRA training #### Key Features for FLUX.1 LoRA training