maxTokensPerDoc property
Per-document token budget; longer documents are truncated (tail
dropped) before scoring. 0 = the SDK default budget. This is the
direct knob on peak memory and per-pair latency on device.
Industry name (Cohere v2 / vLLM max_tokens_per_doc).
Implementation
@$pb.TagNumber(2)
$core.int get maxTokensPerDoc => $_getIZ(1);
Implementation
@$pb.TagNumber(2)
set maxTokensPerDoc($core.int value) => $_setUnsignedInt32(1, value);