Commit
•
4cf3f6c
1
Parent(s):
499793c
Update README.md
Browse files
README.md
CHANGED
|
@@ -27,7 +27,15 @@ Model can generate Greek text with Greek reference speech, English text with Eng
|
|
| 27 |
- Greek Single Speaker Speech Dataset (https://www.kaggle.com/datasets/bryanpark/greek-single-speaker-speech-dataset)
|
| 28 |
- Small part of Emilia Dataset (https://huggingface.co/datasets/amphion/Emilia-Dataset) (EN-B000049.tar)
|
| 29 |
|
| 30 |
-
## Training
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
- Learning Rate: 0.00001
|
| 33 |
- Batch Size per GPU: 3200
|
|
|
|
| 27 |
- Greek Single Speaker Speech Dataset (https://www.kaggle.com/datasets/bryanpark/greek-single-speaker-speech-dataset)
|
| 28 |
- Small part of Emilia Dataset (https://huggingface.co/datasets/amphion/Emilia-Dataset) (EN-B000049.tar)
|
| 29 |
|
| 30 |
+
## Training
|
| 31 |
+
|
| 32 |
+
Training was done in a single RTX 3090.
|
| 33 |
+
|
| 34 |
+
After some manual evaluation two checkpoints produced better results:
|
| 35 |
+
- 225K steps (TBA)
|
| 36 |
+
- 325K steps (TBA)
|
| 37 |
+
|
| 38 |
+
### Arguments
|
| 39 |
|
| 40 |
- Learning Rate: 0.00001
|
| 41 |
- Batch Size per GPU: 3200
|