Attention

Inspect one Llama decoder block with per-head attention and residual-stream math

Loading Attention...
0.0 ms