chunked/content_aware_chunking/_generation_strategies/chunk_31.txt

The main
model then validates the candidate tokens in a single forward pass, which speeds up the decoding process.