Skip to content

server : fix usage.completion_tokens undercount with n > 1 #29486

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ngxson merged 1 commit into from
Oct 20, 2026
Merged

server : fix usage.completion_tokens undercount with n > 1 #29486

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions tools/server/server-context.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4367,10 +4367,20 @@ std::unique_ptr<server_res_generator> server_routes::handle_completions_impl(
res->ok(arr[0]);
} else if (res_type == TASK_RESPONSE_TYPE_OAI_CHAT || res_type == TASK_RESPONSE_TYPE_OAI_CMPL) {
// if multiple results in OAI format, we need to re-format them
json & usage = arr[0]["usage"];
json & choices = arr[0]["choices"];
for (size_t i = 1; i < arr.size(); i++) {
choices.push_back(std::move(arr[i]["choices"][0]));
}

// OAI counts generated tokens across all choices
int64_t n_completion_tokens = 0;
for (auto & item : arr) {
n_completion_tokens += item["usage"]["completion_tokens"].get<int64_t>();
}
usage["completion_tokens"] = n_completion_tokens;
usage["total_tokens"] = usage["prompt_tokens"].get<int64_t>() + n_completion_tokens;

res->ok(arr[0]);
} else {
// multi-results, non-OAI compat
Expand Down
4 changes: 4 additions & 0 deletions tools/server/tests/unit/test_chat_completion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -587,6 +587,10 @@ def test_chat_completions_multiple_choices():
for choice in res.body["choices"]:
assert "assistant" == choice["message"]["role"]
assert choice["finish_reason"] == "length"
# each choice generates max_tokens tokens, usage counts them all
# ref: https://github.com/ggml-org/llama.cpp/issues/29451
assert res.body["usage"]["completion_tokens"] == 16
assert res.body["usage"]["total_tokens"] == res.body["usage"]["prompt_tokens"] + 16


def test_chat_completions_token_count():
Expand Down
Loading