fef18b60bc
Refactor E2E CI to make it clear and faster
1. remove some uesless e2e test
2. remove some uesless function
3. Make sure all test runs with VLLMRunner to avoid oom error
4. Make sure all ops test end with torch.empty_cache to avoid oom error
5. run the test one by one to avoid resource limit error
- vLLM version: v0.10.1.1
- vLLM main:
a344a5aa0a
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
24 lines
984 B
Python
24 lines
984 B
Python
import pytest
|
|
from modelscope import snapshot_download # type: ignore
|
|
|
|
from tests.e2e.conftest import VllmRunner
|
|
from tests.e2e.singlecard.test_ilama_lora import (EXPECTED_LORA_OUTPUT,
|
|
MODEL_PATH, do_sample)
|
|
|
|
|
|
@pytest.mark.parametrize("distributed_executor_backend", ["mp"])
|
|
def test_ilama_lora_tp2(distributed_executor_backend, ilama_lora_files):
|
|
with VllmRunner(snapshot_download(MODEL_PATH),
|
|
enable_lora=True,
|
|
max_loras=4,
|
|
dtype="half",
|
|
max_model_len=1024,
|
|
max_num_seqs=16,
|
|
tensor_parallel_size=2,
|
|
distributed_executor_backend=distributed_executor_backend,
|
|
enforce_eager=True) as vllm_model:
|
|
output = do_sample(vllm_model.model, ilama_lora_files, lora_id=2)
|
|
|
|
for i in range(len(EXPECTED_LORA_OUTPUT)):
|
|
assert output[i] == EXPECTED_LORA_OUTPUT[i]
|