madhushripatil032003
/

readnet-onnx

Automatic Speech Recognition

Model card Files Files and versions

Vakyansh Hindi wav2vec2, as ONNX (for ReadNet in the browser)

An ONNX export of Harveenchadha/vakyansh-wav2vec2-hindi-him-4200 by Harveen Chadha (MIT licence), unchanged apart from the format: full precision, with the output passed through log-softmax.

hi.onnx: input input_values (1 × samples, 16 kHz mono, normalised to zero mean and unit variance); output log_probs (1 × frames × 67).
vocab_hi.json: the CTC vocabulary. The blank is <s> (id 0).

Used by ReadNet (https://github.com/patilllmadhushri-ship-it/Evals, readnet/) to transcribe children's reading and score pronunciation (GOP) in the browser.

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for madhushripatil032003/readnet-onnx

Base model

Harveenchadha/vakyansh-wav2vec2-hindi-him-4200

Quantized

(1)

this model