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

docs typo

This commit is contained in:
Damian Stewart 2023-05-14 12:09:34 +02:00 committed by Victor Hall
parent 1b432d8819
commit a0fca99972
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
doc/OPTIMIZER.md
View File

@ -71,13 +71,13 @@ The SD2.1 text encoder is arranged as follows:
embeddings -> CLIP text encoder (23 layers) -> final layer norm embeddings -> CLIP text encoder (23 layers) -> final layer norm
``` ```
(The SD1.5 text encoder is similar but it has only 12 CLIP layers.) Typically you would apply freezing starting from the left and moving to the right (although it might be interesting to experiment with different freezing patterns). You can control this using the following parameters: (The SD1.5 text encoder is similar but it has only 12 CLIP layers.) Typically you would apply freezing starting from the left and moving to the right, although it might be interesting to experiment with different freezing patterns. You can control this using the following parameters:
* `freeze_embeddings` freezes the front 2 layers (the text embeddings - recommend). * `freeze_embeddings` freezes the front 2 layers (the text embeddings - recommend).
* `freeze_front_n_layers` freezes the front N layers of the CLIP text encoder. You can also pass null to leave the CLIP layers unfrozen, or negative values to count from the back - in the example above `-6` will freeze all but the last 6 layers. * `freeze_front_n_layers` freezes the front N layers of the CLIP text encoder. You can also pass null to leave the CLIP layers unfrozen, or negative values to count from the back. In the example above, `-6` will freeze all but the last 6 layers.
* `freeze_final_layer_norm` freezes the parameters for the text encoder's final `LayerNorm` operation. * `freeze_final_layer_norm` freezes the parameters for the text encoder's final `LayerNorm` operation.
Recommended settings for SD2.1 are provided in `optimizerSD21.json` - frozen embeddings, all CLIP layers frozen except for the last 6, final layer norm unfrozen. If you want to experiment, start by trying different values for `freeze_front_n_layers` - -2 is slower but seems to produce higher quality, whereas -10 is faster but can be more difficult to control. Recommended settings for SD2.1 are provided in `optimizerSD21.json`: frozen embeddings, all CLIP layers frozen except for the last 6, final layer norm unfrozen. If you want to experiment, start by trying different values for `freeze_front_n_layers`: `-2` is slower but seems to produce a higher quality model, whereas `-10` is faster but can be more difficult to control.
## General Beta, weight decay, epsilon, etc tuning ## General Beta, weight decay, epsilon, etc tuning