|
|
|
|
|
|
by Scene_Cast2
310 days ago
|
|
|
Should be in the RoPE paper. The OG transformers used multiplicative sinusoidal embeddings, while RoPE does a pairwise rotation. There's also NoPE, I think SmolLM3 "uses NoPE" (aka doesn't use any positional stuff) every fourth layer. |
|
|