tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 03:55:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

server: fix/test add_generation_prompt (#13770)

Browse Source 
Co-authored-by: ochafik <ochafik@google.com>
This commit is contained in:
Olivier Chafik
2025-05-25 10:45:49 +01:00
committed by GitHub
parent 4032ca4066
commit d785f9c1fd
2 changed files with 26 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
tools/server/tests/unit/test_template.py
View File

@ -47,3 +47,28 @@ def test_date_inside_prompt(template_name: str, format: str, tools: list[dict]):
today_str = datetime.date.today().strftime(format)
assert today_str in prompt, f"Expected today's date ({today_str}) in content ({prompt})"
@pytest.mark.parametrize("add_generation_prompt", [False, True])
@pytest.mark.parametrize("template_name,expected_generation_prompt", [
("meta-llama-Llama-3.3-70B-Instruct", "<|start_header_id|>assistant<|end_header_id|>"),
])
def test_add_generation_prompt(template_name: str, expected_generation_prompt: str, add_generation_prompt: bool):
global server
server.jinja = True
server.chat_template_file = f'../../../models/templates/{template_name}.jinja'
server.start(timeout_seconds=TIMEOUT_SERVER_START)
res = server.make_request("POST", "/apply-template", data={
"messages": [
{"role": "user", "content": "What is today?"},
],
"add_generation_prompt": add_generation_prompt,
})
assert res.status_code == 200
prompt = res.body["prompt"]
if add_generation_prompt:
assert expected_generation_prompt in prompt, f"Expected generation prompt ({expected_generation_prompt}) in content ({prompt})"
else:
assert expected_generation_prompt not in prompt, f"Did not expect generation prompt ({expected_generation_prompt}) in content ({prompt})"

1
tools/server/utils.hpp
View File

@ -731,6 +731,7 @@ static json oaicompat_chat_params_parse(
inputs.grammar = grammar;
inputs.use_jinja = opt.use_jinja;
inputs.parallel_tool_calls = json_value(body, "parallel_tool_calls", false);
inputs.add_generation_prompt = json_value(body, "add_generation_prompt", true);
inputs.reasoning_format = opt.reasoning_format;
if (!inputs.tools.empty() && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE && body.contains("grammar")) {
throw std::runtime_error("Cannot use custom grammar constraints with tools.");