← Back to context Comment by ReptileMan 16 hours ago We have the weights, we have the papers, nothing preventing development on this foundation. 4 comments ReptileMan Reply hgoel 13 hours ago Except compute ReptileMan 12 hours ago Do you think we are near the theoretical maximums of how much model we cram into 20B? hgoel 11 hours ago I don't consider myself qualified to make that call, though I believe there's plenty of room for improvement. But I don't mean inference compute, which is fairly accessible for well-off enthusiasts even now. Training is a lot more expensive though. 1 reply →
hgoel 13 hours ago Except compute ReptileMan 12 hours ago Do you think we are near the theoretical maximums of how much model we cram into 20B? hgoel 11 hours ago I don't consider myself qualified to make that call, though I believe there's plenty of room for improvement. But I don't mean inference compute, which is fairly accessible for well-off enthusiasts even now. Training is a lot more expensive though. 1 reply →
ReptileMan 12 hours ago Do you think we are near the theoretical maximums of how much model we cram into 20B? hgoel 11 hours ago I don't consider myself qualified to make that call, though I believe there's plenty of room for improvement. But I don't mean inference compute, which is fairly accessible for well-off enthusiasts even now. Training is a lot more expensive though. 1 reply →
hgoel 11 hours ago I don't consider myself qualified to make that call, though I believe there's plenty of room for improvement. But I don't mean inference compute, which is fairly accessible for well-off enthusiasts even now. Training is a lot more expensive though. 1 reply →
Except compute
Do you think we are near the theoretical maximums of how much model we cram into 20B?
I don't consider myself qualified to make that call, though I believe there's plenty of room for improvement. But I don't mean inference compute, which is fairly accessible for well-off enthusiasts even now. Training is a lot more expensive though.
1 reply →