After tokenization, each piece becomes a list of numbers so that “skiff” and “dinghy” land nearer than “skiff” and “invoice.” What are those vectors?
Select an answer to reveal the explanation.
Short Explanation
After the split, each token becomes a list of numbers so “skiff” sits nearer “dinghy” than “invoice.” That list is an embedding, a learned meaning vector, not the raw token id. A context window is a length limit, and NCCL ranks are cluster seats, not word neighbors.
Full Explanation
An embedding is a learned numeric meaning vector. The raw token id is just an index. A context window is a length limit, not a vector, and NCCL ranks are out of scope.