ggml-org/llama.cpp

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

Fechada

#6.263 aberto em 23 de mar. de 2024

 (8 comentários) (1 reação) (0 responsável)C++ (21.812 forks)batch import
enhancementgood first issuehelp wanted

Métricas do repositório

Stars
 (124.339 estrelas)
Métricas de merge de PR
 (Mesclagem média 6d 8h) (389 fundiu PRs em 30d)

Description

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.

Guia do colaborador