Tunar-stt-tiny / README.md
openlimi's picture
Upload Tunar-stt-tiny model from Google Drive
ebbfa47 verified
|
Raw History Blame Contribute Delete
1.14 kB
metadata
language:
  - uz
library_name: transformers
base_model: moonshine-ai/moonshine-tiny
tags:
  - automatic-speech-recognition
  - moonshine
  - uzbek
pipeline_tag: automatic-speech-recognition

Tunar-stt-tiny

Uzbek speech recognition model fine-tuned from Moonshine-tiny (v3_1500h run).

  • Output alphabet: new Uzbek Latin alphabet (ö ğ ş ç); convert with o'->ö, g'->ğ, sh->ş, ch->ç.
  • Custom 4096-token BPE tokenizer.
  • Input: 16 kHz mono audio, up to ~30 s per segment.

Usage

import torch, soundfile as sf
from transformers import MoonshineForConditionalGeneration, AutoTokenizer, AutoFeatureExtractor

repo = "openlimi/Tunar-stt-tiny"
model = MoonshineForConditionalGeneration.from_pretrained(repo).eval()
tok = AutoTokenizer.from_pretrained(repo)
fe = AutoFeatureExtractor.from_pretrained(repo)

wav, sr = sf.read("audio.wav")          # 16 kHz mono
inputs = fe(wav, sampling_rate=16000, return_tensors="pt", return_attention_mask=True)
with torch.no_grad():
    ids = model.generate(**inputs, max_new_tokens=int(len(wav) / 16000 * 12) + 8)
print(tok.batch_decode(ids, skip_special_tokens=True)[0])