Re:Show HN: less than 650 LOC trainable GPT only using NumPy...
by p1esk on Friday November 17, 2023 @02:13PM
(#38308513)
I wonder how easy it would be to port this library from numpy to cupy.
[ Parent
| on HN ]
Re:Show HN: less than 650 LOC trainable GPT only using NumPy...
by eslaught on Friday November 17, 2023 @04:03PM
(#38310283)
Or cuNumeric: https://developer.nvidia.com/cunumeric
Re:Show HN: less than 650 LOC trainable GPT only using NumPy...
by joennlae on Friday November 17, 2023 @06:26PM
(#38312145)
This would be interesting to consider. But at the moment, nothing is optimized, so many things must be tackled first (especially in the backwards path, for example, buffering) to justify moving to cupy. The goal was to use it as an educational exercise for me.