Xenova/tiny-random-WhisperForConditionalGeneration

1---2tags:3- transformers.js4---5 6Code to generate:7 8```py9from transformers import WhisperForConditionalGeneration, AutoProcessor10 11new_config_values = dict(12  d_model = 16,13  decoder_attention_heads = 4,14  decoder_layers = 1,15  encoder_attention_heads = 4,16  encoder_layers = 1,17  num_hidden_layers = 1,18 19  ignore_mismatched_sizes=True,20)21original_model = WhisperForConditionalGeneration.from_pretrained('openai/whisper-tiny', **new_config_values)22original_model.save_pretrained('converted')23 24original_processor = AutoProcessor.from_pretrained('openai/whisper-tiny')25original_processor.save_pretrained('converted')26```27 28Followed by:29```sh30$ mkdir -p ./converted/onnx31$ optimum-cli export onnx -m ./converted ./converted/onnx --task automatic-speech-recognition-with-past32$ find ./converted/onnx -type f ! -name "*.onnx" -delete33```34 35## Usage (Transformers.js)36 37If you haven't already, you can install the [Transformers.js](https://huggingface.co/docs/transformers.js) JavaScript library from [NPM](https://www.npmjs.com/package/@huggingface/transformers) using:38```bash39npm i @huggingface/transformers40```41 42**Example:** Transcribe audio from a URL.43 44```js45import { pipeline } from '@huggingface/transformers';46 47const transcriber = await pipeline('automatic-speech-recognition', 'Xenova/tiny-random-WhisperForConditionalGeneration');48const url = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/jfk.wav';49const output = await transcriber(url);50```51 52Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using [🤗 Optimum](https://huggingface.co/docs/optimum/index) and structuring your repo like this one (with ONNX weights located in a subfolder named `onnx`).