Fine-tuning-ready checkpoint for Sidama (sid), built from facebook/mms-tts-sid for use with the WAXAL TTS dataset and the finetune-hf-vits pipeline.


WAXAL config	`google/WaxalNLP` - `sid_tts`
Data provider	Digital Umuganda
Base model	`facebook/mms-tts-sid`
License	CC-BY-NC 4.0 (inherited from base model)

What this repository adds

facebook/mms-tts-{iso} checkpoints are inference-only releases that crash run_vits_finetuning.py. This repository applies three patches:

File	Change
`config.json`	`pad_token_id` set to `0` (was `null`)
`tokenizer_config.json`	`pad_token` entry added
`preprocessor_config.json`	Added - `VitsFeatureExtractor` config from `ylacombe/mms-tts-eng-train`

Model weights are not stored here. _name_or_path in config.json points to facebook/mms-tts-sid, so run_vits_finetuning.py loads weights from the original Facebook checkpoint at training time.

preprocessor_config.json

Downloaded verbatim from ylacombe/mms-tts-eng-train. Values are VITS architecture constants shared by all MMS-TTS languages.

Field	Value
`feature_extractor_type`	`VitsFeatureExtractor`
`feature_size`	`80`
`hop_length`	`256`
`max_wav_value`	`32768.0`
`n_fft`	`1024`
`padding_side`	`right`
`padding_value`	`0.0`
`return_attention_mask`	`False`
`sampling_rate`	`16000`
`spec_gain`	`1`

Usage in finetune-hf-vits

json

{
  "model_name_or_path":     "rnjema-unima/mms-tts-sid-train",
  "feature_extractor_name": "rnjema-unima/mms-tts-sid-train",
  "dataset_name":           "google/WaxalNLP",
  "dataset_config_name":    "sid_tts",
  "audio_column_name":      "audio",
  "text_column_name":       "text",
  "train_split_name":       "train",
  "eval_split_name":        "validation"
}

Inference (after fine-tuning)

python

from transformers import VitsModel, VitsTokenizer
import torch, scipy

model     = VitsModel.from_pretrained("your-org/your-finetuned-model")
tokenizer = VitsTokenizer.from_pretrained("your-org/your-finetuned-model")

inputs = tokenizer("Your text in Sidama.", return_tensors="pt")
with torch.no_grad():
    out = model(**inputs)

scipy.io.wavfile.write("output.wav", model.config.sampling_rate,
    out.waveform.squeeze().numpy())

Technical details


Architecture	VITS (end-to-end, no separate vocoder)
`pad_token_id`	`0`
`vocab_size`	`38`
`is_uroman`	`false`
`sampling_rate`	`16000` Hz

rnjema-unima/mms-tts-sid-baseline

WAXAL MMS-TTS - Sidama (sid)

What this repository adds

preprocessor_config.json

Usage in finetune-hf-vits

Inference (after fine-tuning)

Technical details

References

rnjema-unima/
mms-tts-sid-baseline

WAXAL MMS-TTS - Sidama (`sid`)