Downloads · 30 days
46
10% of all-time downloads
ysdede/parakeet-tdt-0.6b-v2-onnx-tfjs4
parakeet-tdt-0.6b-v2-onnx-tfjs4 is a automatic speech recognition model from ysdede. Use it when you need speech turned into text. It is set up for transformers.js. The card lists the license as cc-by-4.0.
This repo is packaged for the custom nemo-conformer-tdt implementation in Transformers.js v4.
Downloads · 30 days
46
10% of all-time downloads
All-time downloads
466
Public
Repo size
4.5 GB
Likes
3
Public
Click a slice to open those files.
.data2.4 GB · 35%
From the Hugging Face model README
This repo is packaged for the custom nemo-conformer-tdt implementation in Transformers.js v4.
Repository root:
config.json (with model_type: "nemo-conformer-tdt" and transformers.js_config.transducer)tokenizer.jsontokenizer_config.jsonspecial_tokens_map.jsonpreprocessor_config.jsonONNX subfolder:
onnx/encoder_model.onnxonnx/decoder_model_merged.onnx*_fp16.onnx, *_int8.onnxonnx/encoder-model.onnx.data is the external data file referenced by onnx/encoder_model.onnx.vocab.txt is kept for provenance/conversion, but runtime uses tokenizer.json + tokenizer_config.json.nemo128.onnx at repo root is optional and currently not used by the current Transformers.js Nemo feature extractor path (JS mel extraction via preprocessor_config.json). Keeping it at root is fine.This packaging was validated with:
AutoConfig.from_pretrained(...)AutoTokenizer.from_pretrained(...)AutoProcessor.from_pretrained(...)NemoConformerForTDT.from_pretrained(...)pipeline('automatic-speech-recognition', ...)Source baseline: istupakov/parakeet-tdt-0.6b-v2-onnx
See:
ATTRIBUTION.mdprovenance/manifest.json