fix(_internals): use n_tokens0 offset when enabling last-token logits in add_sequence - #2205

Merged

abetlen merged 3 commits into

abetlen:mainfrom

Anai-Guo:fix-add-sequence-logits-index

May 4, 2026

Commits on May 3, 2026

fix(_internals): correct logits index in LlamaBatch.add_sequence for multi-sequence batches
Anai-Guo
committed
6ff43ba

Commits on May 4, 2026

docs: update changelog for batched embeddings fix
abetlen
committed
a3e35c4
test: cover batched embedding reuse semantics
abetlen
committed
051ecf9