Parameter-space extrapolation shaved five steps off the previous record.
N.B. p-values are as follows: P(invalid) = 2.3e-5 P(non-improvement) = 0.168
New Modded-NanoGPT optimization SOTA: @tensorqt has achieved a 2925-step run (-5 steps vs. prev SOTA) by adding a late parameter-space extrapolation step to the previous record.
No Digg Deeper questions have been answered for this story yet.
New Modded-NanoGPT optimization SOTA: @tensorqt has achieved a 2925-step run (-5 steps vs. prev SOTA) by adding a late parameter-space extrapolation step to the previous record.
@kellerjordan0 @tensorqt yoooooo the lich 🐐
New Modded-NanoGPT optimization SOTA: @tensorqt has achieved a 2925-step run (-5 steps vs. prev SOTA) by adding a late parameter-space extrapolation step to the previous record.