root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-13 20:14:29 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

server : try to fix infill when prompt is empty

Browse Source
This commit is contained in:
Georgi Gerganov 2024-01-09 11:27:29 +02:00
parent 8c58330318
commit 24096933b0
No known key found for this signature in database
GPG Key ID: BF970631944C16B7
1 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
examples/server/server.cpp
View File

@ -1399,7 +1399,7 @@ struct llama_server_context
task.multitask_id = multitask_id;
// when a completion task's prompt array is not a singleton, we split it into multiple requests
if (task.data.at("prompt").size() > 1)
if (task.data.count("prompt") && task.data.at("prompt").size() > 1)
{
lock.unlock(); // entering new func scope
return split_multiprompt_task(task);
@ -1570,9 +1570,9 @@ struct llama_server_context
slot->reset();
slot->infill = task.infill_mode;
slot->embedding = task.embedding_mode;
slot->task_id = task.id;
slot->infill = task.infill_mode;
slot->embedding = task.embedding_mode;
slot->task_id = task.id;
slot->multitask_id = task.multitask_id;
if (!launch_slot_with_data(slot, task.data))
@ -1724,7 +1724,8 @@ struct llama_server_context
const bool has_prompt = slot.prompt.is_array() || (slot.prompt.is_string() && !slot.prompt.get<std::string>().empty()) || !slot.images.empty();
// empty prompt passed -> release the slot and send empty response
if (slot.state == IDLE && slot.command == LOAD_PROMPT && !has_prompt)
// note: infill mode allows empty prompt
if (slot.state == IDLE && slot.command == LOAD_PROMPT && !has_prompt && !slot.infill)
{
slot.release();
slot.print_timings();
@ -2581,8 +2582,8 @@ static json format_final_response_oaicompat(const json &request, const task_resu
{"object", streaming ? "chat.completion.chunk" : "chat.completion"},
{"usage",
json{{"completion_tokens", num_tokens_predicted},
{"prompt_tokens", num_prompt_tokens},
{"total_tokens", num_tokens_predicted + num_prompt_tokens}}},
{"prompt_tokens", num_prompt_tokens},
{"total_tokens", num_tokens_predicted + num_prompt_tokens}}},
{"id", gen_chatcmplid()}};
if (server_verbose) {