This model transcribes speech in lowercase Ukrainian alphabet including spaces and apostrophes, and is trained on 69 hours of Ukrainian speech data. It is a non-autoregressive "large" variant of Streaming Citrinet, with around 141 million parameters. Model is fine-tuned from pre-trained Russian Citrinet-1024 model on Ukrainian speech data using Cross-Language Transfer Learning  approach. See the model architecture section and NeMo documentation for complete architecture details. It is also compatible with NVIDIA Riva for production-grade server deployments.
The model is available for use in the NeMo toolkit , and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
To train, fine-tune or play with the model you will need to install NVIDIA NeMo. We recommend you install it after you've installed the latest PyTorch version.
pip install nemo_toolkit['all']
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.EncDecCTCModel.from_pretrained("nvidia/stt_uk_citrinet_1024_gamma_0_25")
First, let's get a sample.
Then simply do:
asr_model.transcribe(['<Path of audio file(s)>'])
python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py pretrained_name="nvidia/stt_uk_citrinet_1024_gamma_0_25" audio_dir="<DIRECTORY CONTAINING AUDIO FILES>"
This model accepts 16000 kHz Mono-channel Audio (wav files) as input.
This model provides transcribed speech as a string for a given audio sample.
Streaming Citrinet-1024 model is a non-autoregressive, streaming variant of Citrinet model  for Automatic Speech Recognition which uses CTC loss/decoding instead of Transducer. You may find more info on this model here: Citrinet Model.
The tokenizer for this models was built using the text transcripts of the train set with this script.
For details on Cross-Lingual transfer learning see .
This model has been trained using validated Mozilla Common Voice Corpus 10.0 dataset (excluding dev and test data) comprising of 69 hours of Ukrainian speech. The Russian model from which this model is fine-tuned has been trained on the union of: (1) Mozilla Common Voice (V7 Ru), (2) Ru LibriSpeech (RuLS), (3) Sber GOLOS and (4) SOVA datasets.
The list of the available models in this collection is shown in the following table. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding.
|Version||Tokenizer||Vocabulary Size||MCV-10 test||MCV-10 dev||MCV-9 test||MCV-9 dev||MCV-8 test||MCV-8 dev|
Since this model was trained on publicly available speech datasets, the performance of this model might degrade for speech that includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.
For the best real-time accuracy, latency, and throughput, deploy the model with NVIDIA Riva, an accelerated speech AI SDK deployable on-prem, in all clouds, multi-cloud, hybrid, at the edge, and embedded. Additionally, Riva provides:
- World-class out-of-the-box accuracy for the most common languages with model checkpoints trained on proprietary data with hundreds of thousands of GPU-compute hours
- Best in class accuracy with run-time word boosting (e.g., brand and product names) and customization of acoustic model, language model, and inverse text normalization
- Streaming speech recognition, Kubernetes compatible scaling, and Enterprise-grade support Check out Riva live demo.
 Citrinet: Closing the Gap between Non-Autoregressive and Autoregressive End-to-End Models for Automatic Speech Recognition
 Google Sentencepiece Tokenizer
 NVIDIA NeMo Toolkit
 Cross-Language Transfer Learning
- Downloads last month
Dataset used to train nvidia/stt_uk_citrinet_1024_gamma_0_25
- Test WER on Mozilla Common Voice 10.0self-reported5.020
- Test WER on Mozilla Common Voice 10.0self-reported4.650
- Test WER on Mozilla Common Voice 9.0self-reported3.750
- Test WER on Mozilla Common Voice 9.0self-reported4.880
- Test WER on Mozilla Common Voice 8.0self-reported3.520
- Test WER on Mozilla Common Voice 8.0self-reported5.020