smajji/nemotron-hinglish-v4

Nemotron Hinglish v4

Nemotron-Hinglish-v4 is a fine-tune of nvidia/nemotron-3.5-asr-streaming-0.6b specialized for English, Hindi and Hinglish (Hindi-English code-switching), built for cache-aware streaming ASR, with improved number / phone-number / digit accuracy.

It preserves the base model's FastConformer-Transducer (RNNT) cache-aware streaming architecture (24 layers, 1024 hidden, 600M params) and its multi-lingual 13088 BPE vocab + 128 language prompts, including the auto language-detection prompt.

Model Details

• Base model: nvidia/nemotron-3.5-asr-streaming-0.6b
• Architecture: FastConformer-Transducer (RNNT), cache-aware streaming, 8x subsampling
• Parameters: ~600M
• Sampling rate: 16 kHz mono
• Target languages: English (en), Hindi (hi), Hinglish code-switched (auto prompt)
• Streaming: cache-aware, chunk sizes 80/160/320/560/1120 ms

Training Data

Fine-tuned on a large bilingual + code-mixed mix with number/symbol-dense corpora added for digit & phone-number accuracy (SPGISpeech, IISc_SPICOR, Peoples Speech, TEDLIUM, earnings22, SPRING Hindi, Shrutilipi, IndicVoices-R, FLEURS en+hi, UJS + Hinglish-CC). Trained with auto language self-detection for code-switching, punctuation/casing preserved.

Benchmark

Greedy decode, punctuation-insensitive WER (held-out sample, auto prompt):

Language	Nemotron-Hinglish-v4
English	4.0%
Hindi	12.0%
Hinglish	29.2%

• Digit / phone-number accuracy improved (fewest digit-sequence errors of any release), from added number-dense corpora (TEDLIUM, earnings22, FLEURS).

Usage (NeMo)

import nemo.collections.asr as nemo_asr
model = nemo_asr.models.ASRModel.restore_from("nvidia/nemotron-hinglish-v4")
model.eval()
transcriptions = model.transcribe(["audio.wav"], batch_size=4)
print(transcriptions)