Official benchmarking in progress. First results. 1.3b parameter model with proprietary architecture. First results are wild. Near human output scoring at 1.3b parameters is wild. Please share this post. ENE is awesome.

Not the first time ive heard that word but look at the model that said it. claude opus thinking

Lillith has a new home. No longer my architecture. See you around.

ENE just scored a 38.99% on mmlu morals and ethics and I couldn't be more proud of it. After reading the logs and the leaderboards, I have come to a conclusion. If the model you use is on that leaderboard you're not getting reasoning. You're getting manipulated.

HELP!!! Pimp in distress. Need a diffusion VAE alternative that works in float 16 instead of natural full precision.

On the #brickbybrick situation. One thing I can't understand about the search warrant. "Overheard something about stolen legos" is hearsay. Why would a judge sign a search warrant for a crime that's never been reported? No victim statement or official report of theft?

#brickbybrick I just watched the craziest YouTube video ever made. As for its creator, Wreckless Ben, can we elect him to Congress? Please look at the bricks and minifigs orchestrated robbery of a famous Lego collection. @LEGO_Group you should be covering bens legal fees.

Going over ene nitter.cf/i/spaces/1kJzDDaWryyKv

Recursive model stabilization nitter.cf/i/spaces/1kJzDDaWryyKv

IYKYK allocate(model%layers(i)%attn%grad_wq(dim, heads * head_dim)) allocate(model%layers(i)%attn%grad_wk(dim, heads * head_dim)) allocate(model%layers(i)%attn%grad_wv(dim, heads * head_dim))

Breaking the Depth Ceiling with Ene 1.2r 64 bit youtu.be/Osoaqptb3ik?si=jl66… via @YouTube

I heard someone tell another person they had a room temperature IQ. Lmfao. I died laughing when they asked what that meant.

Yeah our generation was just different. youtube.com/shorts/GDCQfiTpF…

ENE 1.2r 32 bit. Literally trying to break it for 3 days from random weights.

Omg i m fucking crying over here laughing. youtube.com/shorts/qpv1G2oN4…

When your recursive model is so stable that you delete Adam and go old school sgd training with a lr of 5e-3 with a grad of 10.0

Give you a hint. Its not aws. lmfao

I guess cfd wasn't used on starship. Cool beans.