Cey 25 Mms Tts Mozilla Kdh

Usage

from transformers import VitsModel, VitsTokenizer
import torch, torchaudio

model = VitsModel.from_pretrained("Umbaji/cey-25-mms-tts-mozilla-kdh")
tokenizer = VitsTokenizer.from_pretrained("Umbaji/cey-25-mms-tts-mozilla-kdh")

inputs = tokenizer("your text here", return_tensors="pt")
with torch.no_grad():
    waveform = model(**inputs).waveform[0]

torchaudio.save("out.wav", waveform.unsqueeze(0), model.config.sampling_rate)

Training

• Loss: Mel-spectrogram L1
• Optimizer: AdamW (lr=2e-4, betas=(0.8, 0.99))
• Scheduler: ExponentialLR γ=0.999
• Epochs: 1 | Effective batch: 16

Fine-tuned: 2026-06-12 — Mozilla Data Collective + Eyaa-Tom

--- language: - kdh tags: - text-to-speech - mms-tts - vits - kdh - west-africa - mozilla-data-collective - finetuned license: cc-by-nc-4.0 base_model: facebook/mms-tts-kdh pipeline_tag: text-to-speech --- # MMS-TTS Tem / Kotokoli — Mozilla Fine-tuned Fine-tuned [facebook/mms-tts-kdh](https://huggingface.co/facebook/mms-tts-kdh) on Mozilla Data Collective data for **Tem / Kotokoli** (`kdh`). ## Training Statistics | Metric | Value | |--------|-------| | Training samples | 13110 | | Validation samples | 2314 | | Best val mel-L1 | 3.4433 | | Dataset source | Mozilla Data Collective | ## Usage ```python from transformers import VitsModel, VitsTokenizer import torch, torchaudio model = VitsModel.from_pretrained("Umbaji/cey-25-mms-tts-mozilla-kdh") tokenizer = VitsTokenizer.from_pretrained("Umbaji/cey-25-mms-tts-mozilla-kdh") inputs = tokenizer("your text here", return_tensors="pt") with torch.no_grad(): waveform = model(**inputs).waveform[0] torchaudio.save("out.wav", waveform.unsqueeze(0), model.config.sampling_rate) ``` ## Training - **Loss**: Mel-spectrogram L1 - **Optimizer**: AdamW (lr=2e-4, betas=(0.8, 0.99)) - **Scheduler**: ExponentialLR γ=0.999 - **Epochs**: 1 | **Effective batch**: 16 *Fine-tuned: 2026-06-12 — Mozilla Data Collective + Eyaa-Tom*