Update README.md
Browse files
README.md
CHANGED
@@ -10,7 +10,7 @@ Code is very simple, add noise during the training loop between the text_encoder
|
|
10 |
perturbation_delta = torch.randn_like(encoder_hidden_states) * (perturbation_deviation)
|
11 |
encoder_hidden_states = encoder_hidden_states + perturbation_delta
|
12 |
|
13 |
-
embedding_perturbation is the tunable
|
14 |
|
15 |
Models are SD1.5 trained for 30 epochs, Unet only, AdamW8bit, 2e-6 constant LR, batch size 12. EveryDream2Trainer config json included.
|
16 |
|
|
|
10 |
perturbation_delta = torch.randn_like(encoder_hidden_states) * (perturbation_deviation)
|
11 |
encoder_hidden_states = encoder_hidden_states + perturbation_delta
|
12 |
|
13 |
+
"embedding_perturbation" is the tunable gamma from the paper.
|
14 |
|
15 |
Models are SD1.5 trained for 30 epochs, Unet only, AdamW8bit, 2e-6 constant LR, batch size 12. EveryDream2Trainer config json included.
|
16 |
|