EveryDream2trainer/doc/OPTIMIZER.md

41 lines
2.5 KiB
Markdown
Raw Normal View History

2023-02-25 14:19:23 -07:00
# Advanced optimizer tweaking
You can set advanced optimizer settings using this arg:
--optimizer_config optimizer.json
or in train.json
"optimizer_config": "optimizer.json"
2023-02-28 19:19:52 -07:00
A default `optimizer.json` is supplied which you can modify.
2023-02-25 14:19:23 -07:00
This has expanded tweaking. This doc is incomplete, but there is information on the web on betas and weight decay setting you can search for.
If you do not set optimizer_config, the defaults are `adamw8bit` with standard betas of `(0.9,0.999)`, weight decay `0.01`, and epsilon `1e-8`. The hyperparameters are originally from XavierXiao's Dreambooth code and based off Compvis Stable Diffusion code.
## Optimizers
In `optimizer.json` the `optimizer` value is the type of optimizer to use. Below are the supported optimizers.
* adamw
Standard full precision AdamW optimizer exposed by PyTorch. Not recommended. Slower and uses more memory than adamw8bit. Widely documented on the web.
* adamw8bit
Tim Dettmers / bitsandbytes AdamW 8bit optimizer. This is the default and recommended setting. Widely documented on the web.
* lion
Lucidrains' [implementation](https://github.com/lucidrains/lion-pytorch) of the [lion optimizer](https://arxiv.org/abs/2302.06675). Click links to read more. Unknown what hyperparameters will work well, but paper shows potentially quicker learning. *Highly experimental, but tested and works.*
## Optimizer parameters
LR can be set in `optimizer.json` and excluded from the main CLI arg or train.json but if you use the main CLI arg or set it in the main train.json it will override the setting. This was done to make sure existing behavior will not break. To set LR in the `optimizer.json` make sure to delete `"lr": 1.3e-6` in your main train.json and exclude the CLI arg.
2023-03-02 14:45:22 -07:00
The text encoder LR can run at a different value to the Unet LR. This may help prevent over-fitting, especially if you're training from SD2 checkpoints. To set the text encoder LR, add a value for `text_encoder_lr_scale` to `optimizer.json`. For example, to train the text encoder with an LR that is half that of the Unet, add `"text_encoder_lr_scale": 0.5` to `optimizer.json`. The default value is `1.0`, meaning the text encoder and Unet are trained with the same LR.
2023-02-25 14:19:23 -07:00
Betas, weight decay, and epsilon are documented in the [AdamW paper](https://arxiv.org/abs/1711.05101) and there is a wealth of information on the web, but consider those experimental to tweak. I cannot provide advice on what might be useful to tweak here.
Note `lion` does not use epsilon.