Amit Moryossef

Model exists, thanks to Zifan. Not yet implemented, waiting for transcription models

Screenshots are now taken via the `npm run mobile:metadata` command, using playwright

Firebase ML has two issues: 1. Does not allow large files 2. Not available for the web  Alternative solution: Cloud storage, upload h5 file, auto converts to coreml and...

For CoreML, the convolutions might be too much: > 2022-07-02 22:57:13.178238+0200 benchmark[20833:399499] Error: Convolution configuration cannot fit in KMEM (Given: 73728b, Max: 65536b)

Proposal: - [x] upload model files to storage. - [x] automatically convert to tflite and coreml files - [x] create a `assets` module, that can download and store models in...

Native build bugs: - [x] Speech to Text - [x] Share video file on iOS to WhatApp - [x] Extra scrolling on iOS - [ ] Avatar augmented reality not...

Another way to determine if to remove the sentence or not would be to remove all sentence translated to exactly `max-length` tokens. This is because in some scenarios, the source...

I would really like to see that too. I work on language pairs with no overlap between the src and target character set, and so a separate tokenization model for...

Did you manage to find the entire dataset?

You got my point indeed. Before I try this, a consultation if I may - How does the memory of a transformer encoder-decoder model behave? 1. $O(max(|src|, |trg|)^2)$ 2. $O(max(|src|^2,...