|
Download README.md from StarlexAI/orbit-phone: direct link, hf CLI and curl.
- Browser
- Download file 4.72 kB
-
https://huggingface.co/StarlexAI/orbit-phone/resolve/main/README.md
- Command line
-
hf download hf://StarlexAI/orbit-phone/README.md
-
curl -L -o README.md https://huggingface.co/StarlexAI/orbit-phone/resolve/main/README.md
4.72 kB
| license: apache-2.0 | |
| language: [en, de, fr] | |
| pipeline_tag: automatic-speech-recognition | |
| library_name: pytorch | |
| tags: [asr, speech-recognition, ctc, conformer, muon, starlex, orbit, from-scratch] | |
| datasets: [openslr/librispeech_asr, parler-tts/mls_eng, facebook/multilingual_librispeech, facebook/voxpopuli, espnet/yodas-granary] | |
| # Orbit Phone v1 | |
| **Speech recognition trained from scratch in one night. 2.60 % word error rate on LibriSpeech test-clean.** | |
|  | |
| Orbit Phone is the speech model of the Orbit family by Starlex AI. It is a 223M parameter Conformer with a CTC | |
| head, trained from random weights with no pretrained components, English first plus German and French. Apache-2.0. | |
| | | | | |
| |---|---| | |
| | Parameters | 223M | | |
| | Training | about 12 hours on 3 GPUs, about 230,000 hours of audio processed | | |
| | Languages | English, German, French | | |
| | License | Apache-2.0 | | |
| ## Results | |
| Word error rate, lower is better. Greedy CTC decoding, no language model, text normalized (lowercase, no | |
| punctuation). | |
| | Test set | Orbit Phone | Whisper small (OpenAI, 244M) | Whisper base (OpenAI, 74M) | | |
| |---|---|---|---| | |
| | LibriSpeech test-clean (2,611 utterances) | **2.60 %** | 3.4 % | 5.0 % | | |
| | LibriSpeech test-other (2,932 utterances) | **6.35 %** | 7.6 % | 12.4 % | | |
| | MLS English test (3,769 utterances) | 9.45 % | | | | |
| | VoxPopuli English test (1,830 utterances) | 7.38 % | | | | |
| | MLS German test (3,394 utterances) | 7.62 % | | | | |
| | VoxPopuli German test (1,957 utterances) | 10.27 % | | | | |
| The Whisper numbers are OpenAI's published results. They are for orientation, not a head to head comparison, | |
| because Orbit Phone saw LibriSpeech training data and Whisper is evaluated zero-shot. Clips longer than 30 s were | |
| skipped in our evaluation. | |
| ## Try it | |
| ```bash | |
| pip install torch safetensors tokenizers soundfile numpy | |
| python orbit_phone.py speech.wav | |
| ``` | |
| ```python | |
| from orbit_phone import load_orbit_phone, transcribe | |
| model, tok, feat = load_orbit_phone(".") | |
| print(transcribe("speech.wav", model, tok, feat)) | |
| ``` | |
| Everything needed to run the model is in this repository. No other library or checkpoint is downloaded. | |
| ## How it was built | |
| 1. **Data.** LibriSpeech (960 h), subsets of MLS English and German, VoxPopuli English and German, and subsets of | |
| YODAS-Granary English, German and French. Mix of about 60 % English, 30 % German, 10 % French. | |
| 2. **Tokenizer.** Own BPE vocabulary with 1,024 pieces. | |
| 3. **Training.** Muon (Newton-Schulz orthogonalized momentum) on the 2D block matrices and AdamW on the rest, with | |
| warmup, a constant phase and linear decay over the last 30 % of a fixed time budget. SpecAugment, bf16, | |
| 243,504 steps. | |
| 4. **Fixed batch shapes.** Audio is sorted into 23 length buckets with about one hour of audio per step. Fixed | |
| shapes made training about 50 times faster than variable shapes on our hardware, from 353x to 19,000x real time. | |
| ## Architecture | |
| - Input: 16 kHz audio, 128 log-mel features at 100 Hz, normalized per utterance | |
| - Encoder: 3 strided convolutions (8x down to 12.5 Hz), 16 Conformer blocks (d 768, 12 heads, RoPE self-attention, | |
| depthwise convolution kernel 31, macaron feed-forward) | |
| - Output: CTC over the 1,024 piece vocabulary, intermediate CTC after block 8 during training | |
| - Output text is lowercase without punctuation in v1 | |
| ## Limitations | |
| - Proof of concept after one night of training. Not for safety-critical or legal transcription. | |
| - No casing, punctuation, timestamps or language identification in v1. | |
| - Weaker on rare names and German word endings. No language model in decoding. | |
| - Audio longer than 30 s is cut into fixed 30 s windows, words at the cut can be lost. | |
| ## The Orbit family | |
| | Model | What it does | | |
| |---|---| | |
| | [Orbit-Nano-Compliance-E2B](https://huggingface.co/StarlexAI/Orbit-Nano-Compliance-E2B) | German and English compliance chat, rank 1 of 7 on our compliance benchmark | | |
| | [Orbit-Nano-Compliance-E2B-GGUF](https://huggingface.co/StarlexAI/Orbit-Nano-Compliance-E2B-GGUF) | The same model for laptops | | |
| | [Orbit Phone](https://huggingface.co/StarlexAI/orbit-phone) | Speech recognition | | |
| | [Orbit-Vision](https://huggingface.co/StarlexAI/orbit-vision) | Questions about images, charts and documents (research preview) | | |
| | [Orbit-Image](https://huggingface.co/StarlexAI/orbit-image) | Pixel art from text (research preview) | | |
| ## Data attribution and license | |
| Weights and code Apache-2.0. Training data licenses are LibriSpeech CC-BY-4.0, Multilingual LibriSpeech CC-BY-4.0, | |
| VoxPopuli CC0 and YODAS-Granary CC-BY-3.0. We thank the authors of these datasets. | |
| Built by Starlex AI, a Foundation AI Lab. |