Mirrors
/
EveryDream2trainer
mirror of https://github.com/victorchall/EveryDream2trainer.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

clarify convert doc

This commit is contained in:
Victor Hall 2023-07-07 01:11:59 -04:00
parent 7686bcd66e
commit cb1264fbe9
1 changed files with 14 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
doc/BASEMODELS.md
View File

@ -55,7 +55,7 @@ And for any SD2.1 768 models (uses v2-v yaml and "v_prediction" prediction type)
Note the `v2-inference-v.yaml` and `v_prediction`. This is because the SD2.1 768 model uses a different yaml and prediction type than the SD1.X models. Note the `v2-inference-v.yaml` and `v_prediction`. This is because the SD2.1 768 model uses a different yaml and prediction type than the SD1.X models.
And finally the SD2.0 512 base model (generally not recommended base model): And finally the SD2.0 512 base model (generally not recommended base model, no one tunes this either):
python utils/convert_original_stable_diffusion_to_diffusers.py --scheduler_type ddim ^ python utils/convert_original_stable_diffusion_to_diffusers.py --scheduler_type ddim ^
--original_config_file v2-inference.yaml ^ --original_config_file v2-inference.yaml ^
@ -67,13 +67,22 @@ And finally the SD2.0 512 base model (generally not recommended base model):
If you have other models, you need to know the base model that was used for them, **in particular use the correct yaml (original_config_file) or it will not properly convert.** Make sure to put some sort of name in the dump_path after "ckpt_cache/" so you can reference it later. If you have other models, you need to know the base model that was used for them, **in particular use the correct yaml (original_config_file) or it will not properly convert.** Make sure to put some sort of name in the dump_path after "ckpt_cache/" so you can reference it later.
All of the above is one time. After running, you will use --resume_ckpt and just name the file without "ckpt_cache/" All of the above is one time. After running, you will use `resume_ckpt` and just name the file without "ckpt_cache/"
ex. ex.
```
train.json
{
...
"resume_ckpt": "my_sd15_model",
...
}
```
or using the CLI arg:
python train.py --resume_ckpt "my_sd21_model" ...
python train.py --resume_ckpt "sd_v1-5_vae" ...
python train.py --resume_ckpt "v2-1_768-ema-pruned" ...
python train.py --resume_ckpt "512-base-ema" ...
## Automatic download ## Automatic download