Parakeet Tdt Ctc 110m Coreml

--- license: cc-by-4.0 language: - en metrics: - wer base_model: - nvidia/parakeet-tdt_ctc-110m pipeline_tag: automatic-speech-recognition tags: - automatic-speech-recognition - speech - audio - Transducer - TDT - FastConformer - Conformer - pytorch - NeMo - hf-asr-leaderboard --- # Parakeet-TDT-CTC 110M — CoreML CoreML export of [nvidia/parakeet-tdt_ctc-110m](https://huggingface.co/nvidia/parakeet-tdt_ctc-110m) for on-device speech recognition on Apple Silicon via [FluidAudio](https://github.com/FluidInference/FluidAudio). ## CoreML Components | File | Size | Description | |------|------|-------------| | `Preprocessor.mlmodelc` | 207 MB | Fused mel-spectrogram + FastConformer encoder | | `Decoder.mlmodelc` | 7.5 MB | 1-layer LSTM prediction network | | `JointDecision.mlmodelc` | 2.7 MB | Single-step joint network (token + duration) | | `parakeet_vocab.json` | 18 KB | 1024-token BPE vocabulary | | `config.json` | 2.5 KB | Model metadata and I/O contracts | **Input:** 16 kHz mono audio, fixed 15-second window (240,000 samples). **Output:** Token IDs, probabilities, and TDT duration predictions per encoder frame. ## Performance Benchmarked with FluidAudio CLI on Apple M2 (release build): | Benchmark | WER | |-----------|-----| | LibriSpeech test-clean | **3.0%** | | RTFx (overall) | **102x** real-time | | Peak memory | 0.3 GB | NVIDIA's reference WER (greedy, GPU): | Benchmark | WER | |-----------|-----| | LibriSpeech test-clean | 2.4% | | LibriSpeech test-other | 5.2% | | AMI | 15.88% | | Earnings-22 | 12.42% | | GigaSpeech | 10.52% | | TEDLIUM-v3 | 4.16% | ## Usage with FluidAudio ```bash # Transcribe fluidaudiocli transcribe audio.wav --model-version tdt-ctc-110m # Benchmark fluidaudiocli asr-benchmark --subset test-clean --model-version tdt-ctc-110m ``` Models auto-download from this repo on first use. To pre-fetch: ```bash fluidaudiocli download --model-version tdt-ctc-110m ``` ## Conversion Exported from NeMo using [mobius/models/stt/parakeet-tdt-ctc-110m/coreml/convert-tdt-coreml.py](https://github.com/FluidInference/mobius): - Preprocessor fuses mel-spectrogram extraction and the FastConformer encoder into a single CoreML model - JointDecision is the single-step variant (encoder_step + decoder_step inputs) used by FluidAudio's TDT decoder - All models exported as MLProgram (iOS 17+ / macOS 14+), float32 precision ## References - [Fast Conformer with Linearly Scalable Attention](https://arxiv.org/abs/2305.05084) - [Efficient Sequence Transduction by Jointly Predicting Tokens and Durations](https://arxiv.org/abs/2304.06795) - [NVIDIA NeMo](https://github.com/NVIDIA/NeMo)