ggml-org/llama.cpp

server: exit failure if `--embedding` is set with an incoherent `--ubatch-size`

已关闭

#6,263 创建于 2024年3月23日

 (8 条评论) (1 个反应) (0 位负责人)C++ (21,812 个派生)batch import
enhancementgood first issuehelp wanted

仓库指标

星标
 (124,339 个星标)
PR 合并指标
 (平均合并 6天 8小时) (30 天内合并 389 个 PR)

描述

Context

there is no advantage to increase n_batch above n_ubatch with embeddings models with pooling, because the entire batch must fit in a physical batch (ie. n_ubatch). n_batch is always >= n_ubatch.

Proposition

Exit failure if --embedding is set and --ubatch-size != --batch-size in the server example. Probably also in the retrieval example in #6193.

Aldo probably KV bert.context_size must be taken into account.

贡献者指南