amrb·2 lata temu·discussIt's a red flag that the 1.2bil model has to fit in gpu memory, happy to be provided wrong when the code drops
amrb·2 lata temu·discussCan't say I mastered the concept either, I'm waiting for the code [0] to be release so I can run some head-to-head tests.[0] https://github.com/Aleph-Alpha/trigrams
amrb·2 lata temu·discussAn alternative approache to BPE tokenization https://arxiv.org/abs/2406.19223
amrb·3 lata temu·discussLoved hearing about the history SQLite! I do feel we should capture more of the tech story's, as its hard to learn from mistakes.. if you don't know the history.