Model card: remove the training section
Browse files
README.md
CHANGED
|
@@ -144,14 +144,6 @@ cpp/build/tts_hi --model tts_hi_q8_0.gguf --text "नमस्ते, आप क
|
|
| 144 |
cpp/build/tts_hi --model tts_hi_q8_0.gguf --serve 8090 --page web --threads 4 # web page on http://127.0.0.1:8090
|
| 145 |
```
|
| 146 |
|
| 147 |
-
## Training
|
| 148 |
-
|
| 149 |
-
- **Data:** [Rasa](https://huggingface.co/datasets/ai4bharat/Rasa) by AI4Bharat, Hindi, one female speaker, with
|
| 150 |
-
its emotion and conversational styles (about 11.7k training utterances).
|
| 151 |
-
- **Recipe:** trained from scratch for 100k steps, with recorded phoneme durations (no learned aligner).
|
| 152 |
-
- Losses: multi-resolution STFT + log-mel, plus duration, pitch and energy losses.
|
| 153 |
-
- Waveform critics (multi-period and multi-resolution) from step 8k, at 0.7× the reconstruction gradient.
|
| 154 |
-
|
| 155 |
## License and credits
|
| 156 |
|
| 157 |
The weights are released under **CC BY 4.0**, following the training data:
|
|
|
|
| 144 |
cpp/build/tts_hi --model tts_hi_q8_0.gguf --serve 8090 --page web --threads 4 # web page on http://127.0.0.1:8090
|
| 145 |
```
|
| 146 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
## License and credits
|
| 148 |
|
| 149 |
The weights are released under **CC BY 4.0**, following the training data:
|