Hate to say it, but this was an instant Buy for me. I love experimenting with new modalities and this is a cool idea pre-packaged. If it doesn't work, what is 200 bucks compared to a SWE salary? It would be a wacky novelty memento of the time. If it is a nice boost to my work by... what... 2% - or just good for a chuckle each day - Good, I'm happy.
Feels like a leap. This kind of move was always possible. It's possible China stops publishing their frontier too. US could lock down access to Nvidia latest hw of scale even if you intended to do open models. Then what? Say amd or bust? The best you could do going solo (i.e no nation-state interference) is tiny stuff that you can run on commercial stuff. But that is seriously limited / slow in comparison. You either have to do dumb and fast, or smart and slow IME for these self-hosted things that aren't on the beefy racks.
Ah very interesting. For this one, I had to dig up an older repo I made where I converted WEB translation into one big json file to make it easier to work with. Not sure the formats exposed and how... pliable? they are. (https://github.com/jacksonStone/Bible_as_JSON)
Yeah a drop down list like that would be sweet. I think you'd have to format / find the open source version / translation of that text. Finding a well formatted and readable, public use (not just nkjv for example) was tricky For WEB I had to do a good b it of data massaging to get it in a consumable format..
Yes! looks like yours has more features (I considered generating the audio) FWIW I observed that the embedding gives the most signal when I pass whole paragraphs (for all I know the Quran is in lyrical verse so maybe that doesn't work)
Ah interesting. So I'm using an English translation. Tbh I wish I could do more with the Arabic but the models are not perfect at that. But the idea of giving it whole paragraphs makes a ton of sense.
Tried to generate audio but wasn't satisfied with the interpretation of some words, saying that thou may try again.
