CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-29 08:41:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

server : return HTTP 400 if prompt exceeds context length (#16486)

Browse Source 
In streaming mode when prompt exceeds context length, the server returns
HTTP 200 status code with a JSON error in the body.  This is very
confusing and inconsistent with all other inference engines which return
HTTP 4xx error in this case.

This patch fixes this problem and makes the server return HTTP 400 in
such cases.
This commit is contained in:
Radoslav Gerganov
2025-10-10 17:11:07 +03:00
committed by GitHub
parent cdb6da468c
commit 68ee98ae18
3 changed files with 40 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
tools/server/server.cpp
View File

@@ -3727,7 +3727,7 @@ struct server_context {
}
} else {
if (slot.n_prompt_tokens() >= slot.n_ctx) {
send_error(slot, "the request exceeds the available context size. try increasing the context size or enable context shift", ERROR_TYPE_EXCEED_CONTEXT_SIZE);
send_error(slot, "the request exceeds the available context size, try increasing it", ERROR_TYPE_EXCEED_CONTEXT_SIZE);
slot.release();
continue;
}
@@ -4955,9 +4955,17 @@ int main(int argc, char ** argv) {
// Everything else, including multimodal completions.
inputs = tokenize_input_prompts(ctx_server.vocab, ctx_server.mctx, prompt, true, true);
}
const size_t n_ctx_slot = ctx_server.n_ctx / ctx_server.params_base.n_parallel;
tasks.reserve(inputs.size());
for (size_t i = 0; i < inputs.size(); i++) {
auto n_prompt_tokens = inputs[i].size();
if (n_prompt_tokens >= n_ctx_slot) {
json error_data = format_error_response("the request exceeds the available context size, try increasing it", ERROR_TYPE_EXCEED_CONTEXT_SIZE);
error_data["n_prompt_tokens"] = n_prompt_tokens;
error_data["n_ctx"] = n_ctx_slot;
res_error(res, error_data);
return;
}
server_task task = server_task(type);
task.id = ctx_server.queue_tasks.get_new_id();