Joshua Lochner

Hi there 👋 This is because the model's weren't exported with `output_attentions=True`, and the current version of Transformers.js v3 doesn't yet support this. Your fork looks pretty good so far...

Does this also happen when using a normal web worker? 👀

> Yes, the version I forked actually works with output attentions of the model but sometimes during the inference phase there are predicted chunks with token. I also checked the...

I did a deep dive on this and I think I've now got it working in v3! ```js import { pipeline } from '@xenova/transformers'; const model_id = 'onnx-community/whisper-tiny.en_timestamped'; const transcriber...

Interesting - thanks for the information! @guschmue any idea what's going wrong?

Thanks everyone for your tests! I can confirm that the regex fix by @MatteoFasulo works pretty well, so I think it's a good idea to integrate some adaptation of it...

Commit made [here](https://github.com/xenova/transformers.js/commit/ea732896d7505c597c4568404adfeb27a222b387). It's significantly better now! cc @sanchit-gandhi I think this should be added to transformers too... it makes word-level timestamps far more accurate!

@kyr0 GitHub search doesn't index branches other than main, so you would need to inspect the code directly. For example, the device is set here: https://github.com/xenova/transformers.js/blob/1b4d2428225ef8f63be94bfa38a0d7fd81ac7c0c/src/models.js#L149-L162

Hi there 👋 I've uploaded the tokenizer files to https://huggingface.co/Xenova/gemma-2-tokenizer, and it's already compatible with Transformers.js :) [Demo](https://huggingface.co/spaces/Xenova/the-tokenizer-playground?tokenizer=Xenova%2Fgemma-2-tokenizer):  You can just update the id in your code and it...

Hi there 👋 You can certainly use Transformers.js for that :) I'd recommend checking out our v3/dev branch ([here](https://github.com/xenova/transformers.js/pull/545)), which enables WebGPU support. You can probably use [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct-onnx-web) for this,...