mirror of
https://github.com/ggerganov/llama.cpp
synced 2026-03-04 22:29:35 +01:00
e7c2cf1356
* llama-server: add router multi-model tests (#17704) Add 4 test cases for model router: - test_router_unload_model: explicit model unloading - test_router_models_max_evicts_lru: LRU eviction with --models-max - test_router_no_models_autoload: --no-models-autoload flag behavior - test_router_api_key_required: API key authentication Tests use async model loading with polling and graceful skip when insufficient models available for eviction testing. utils.py changes: - Add models_max, models_dir, no_models_autoload attributes to ServerProcess - Handle JSONDecodeError for non-JSON error responses (fallback to text) * llama-server: update test models to new HF repos * add offline * llama-server: fix router LRU eviction test and add preloading Fix eviction test: load 2 models first, verify state, then load 3rd to trigger eviction. Previous logic loaded all 3 at once, causing first model to be evicted before verification could occur. Add module fixture to preload models via ServerPreset.load_all() and mark test presets as offline to use cached models * llama-server: fix split model download on Windows --------- Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com>
97 lines
2.7 KiB
Python
97 lines
2.7 KiB
Python
import pytest
|
|
import requests
|
|
from utils import *
|
|
|
|
server = ServerPreset.tinyllama2()
|
|
|
|
|
|
@pytest.fixture(autouse=True)
|
|
def create_server():
|
|
global server
|
|
server = ServerPreset.tinyllama2()
|
|
|
|
|
|
def test_server_start_simple():
|
|
global server
|
|
server.start()
|
|
res = server.make_request("GET", "/health")
|
|
assert res.status_code == 200
|
|
|
|
|
|
def test_server_props():
|
|
global server
|
|
server.start()
|
|
res = server.make_request("GET", "/props")
|
|
assert res.status_code == 200
|
|
assert ".gguf" in res.body["model_path"]
|
|
assert res.body["total_slots"] == server.n_slots
|
|
default_val = res.body["default_generation_settings"]
|
|
assert server.n_ctx is not None and server.n_slots is not None
|
|
assert default_val["n_ctx"] == server.n_ctx / server.n_slots
|
|
assert default_val["params"]["seed"] == server.seed
|
|
|
|
|
|
def test_server_models():
|
|
global server
|
|
server.start()
|
|
res = server.make_request("GET", "/models")
|
|
assert res.status_code == 200
|
|
assert len(res.body["data"]) == 1
|
|
assert res.body["data"][0]["id"] == server.model_alias
|
|
|
|
|
|
def test_server_slots():
|
|
global server
|
|
|
|
# without slots endpoint enabled, this should return error
|
|
server.server_slots = False
|
|
server.start()
|
|
res = server.make_request("GET", "/slots")
|
|
assert res.status_code == 501 # ERROR_TYPE_NOT_SUPPORTED
|
|
assert "error" in res.body
|
|
server.stop()
|
|
|
|
# with slots endpoint enabled, this should return slots info
|
|
server.server_slots = True
|
|
server.n_slots = 2
|
|
server.start()
|
|
res = server.make_request("GET", "/slots")
|
|
assert res.status_code == 200
|
|
assert len(res.body) == server.n_slots
|
|
assert server.n_ctx is not None and server.n_slots is not None
|
|
assert res.body[0]["n_ctx"] == server.n_ctx / server.n_slots
|
|
assert "params" not in res.body[0]
|
|
|
|
|
|
def test_load_split_model():
|
|
global server
|
|
server.offline = False
|
|
server.model_hf_repo = "ggml-org/models"
|
|
server.model_hf_file = "tinyllamas/split/stories15M-q8_0-00001-of-00003.gguf"
|
|
server.model_alias = "tinyllama-split"
|
|
server.start()
|
|
res = server.make_request("POST", "/completion", data={
|
|
"n_predict": 16,
|
|
"prompt": "Hello",
|
|
"temperature": 0.0,
|
|
})
|
|
assert res.status_code == 200
|
|
assert match_regex("(little|girl)+", res.body["content"])
|
|
|
|
|
|
def test_no_webui():
|
|
global server
|
|
# default: webui enabled
|
|
server.start()
|
|
url = f"http://{server.server_host}:{server.server_port}"
|
|
res = requests.get(url)
|
|
assert res.status_code == 200
|
|
assert "<!doctype html>" in res.text
|
|
server.stop()
|
|
|
|
# with --no-webui
|
|
server.no_webui = True
|
|
server.start()
|
|
res = requests.get(url)
|
|
assert res.status_code == 404
|