Compare commits

...
Sign in to create a new pull request.

1 commit

Author SHA1 Message Date
Georgi Gerganov
33d7b70c88
server : do not speculate during prompt processing
ggml-ci
2024-12-03 10:58:43 +02:00

View file

@ -2322,6 +2322,10 @@ struct server_context {
continue; continue;
} }
if (slot.state != SLOT_STATE_GENERATING) {
continue;
}
llama_token id = slot.sampled; llama_token id = slot.sampled;
struct common_speculative_params params_spec; struct common_speculative_params params_spec;