Instructions to use Edmon02/speecht5_finetuned_hy with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Edmon02/speecht5_finetuned_hy with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-to-speech", model="Edmon02/speecht5_finetuned_hy")# Load model directly from transformers import AutoProcessor, AutoModelForTextToSpectrogram processor = AutoProcessor.from_pretrained("Edmon02/speecht5_finetuned_hy") model = AutoModelForTextToSpectrogram.from_pretrained("Edmon02/speecht5_finetuned_hy") - Notebooks
- Google Colab
- Kaggle
Archived baseline: speecht5_finetuned_hy
First fine-tune of microsoft/speecht5_tts on Common Voice 11 (multilingual tags: hy, en, nl). Vocab 81.
Status
Archived โ kept for lineage reproducibility. Armenian production work moved to HyVoxPopuli-based checkpoints.
Successor chain
speecht5_finetuned_hy (this repo)
โ speecht5_finetuned_voxpopuli_nl
โ speecht5_finetuned_voxpopuli_hy โ production
โ TTS_NB_2 โ active training
Evaluation (historical)
| Step | Validation loss |
|---|---|
| 1000 | 0.4785 |
Training hyperparameters
- Learning rate: 1e-5
- Effective batch size: 32
- Training steps: 1000
- Transformers 4.35.2, PyTorch 2.1.0
Use instead
| Task | Model |
|---|---|
| Armenian TTS | Edmon02/speecht5_finetuned_voxpopuli_hy |
| Fine-tuning | Edmon02/TTS_NB_2 |
License
MIT
- Downloads last month
- 174
Model tree for Edmon02/speecht5_finetuned_hy
Base model
microsoft/speecht5_tts