Update README.md
Browse files
README.md
CHANGED
|
@@ -27,7 +27,9 @@ Just use natural language,
|
|
| 27 |
the new Lora training is 500 times better than my last method there seems to be no more bleed where the Lora damages the quality of the default model,
|
| 28 |
i think this may be the final version as it does exactly as i expected, it was captioned with a large amount of relevant words to associate them,
|
| 29 |
i have yet to spend the time to work out what is the best epoch so i just uploaded them all try from 15 onwards but the final epoch seems awesome even at 1.0 strength,
|
| 30 |
-
for any wondering the new training method is now using the RAW model and using
|
|
|
|
|
|
|
| 31 |
|
| 32 |
## Trigger words
|
| 33 |
|
|
|
|
| 27 |
the new Lora training is 500 times better than my last method there seems to be no more bleed where the Lora damages the quality of the default model,
|
| 28 |
i think this may be the final version as it does exactly as i expected, it was captioned with a large amount of relevant words to associate them,
|
| 29 |
i have yet to spend the time to work out what is the best epoch so i just uploaded them all try from 15 onwards but the final epoch seems awesome even at 1.0 strength,
|
| 30 |
+
for any wondering the new training method is now using the RAW model and using https://github.com/shootthesound/Fizgig.
|
| 31 |
+
|
| 32 |
+
* after testing will do a v4 removing watermarks
|
| 33 |
|
| 34 |
## Trigger words
|
| 35 |
|