[CI/Build] Reduce LoRA model tests. (#42266)

Signed-off-by: Jee Jee Li <jeejeelee@inferact.ai>
This commit is contained in:
Jee Jee Li
2026-05-11 14:49:08 +08:00
committed by GitHub
parent 581b5e9afc
commit 05d610e5cd
11 changed files with 50 additions and 31 deletions
+4
View File
@@ -10,6 +10,7 @@ import vllm.config
from vllm import LLM
from vllm.lora.request import LoRARequest
from vllm.model_executor.model_loader.tensorizer import TensorizerConfig
from vllm.platforms import current_platform
from ..utils import VLLM_PATH, create_new_process_for_each_test, multi_gpu_test
@@ -139,6 +140,9 @@ def test_llama_lora(llama32_lora_files, cudagraph_specialize_lora: bool):
generate_and_test(llm, llama32_lora_files)
@pytest.mark.skipif(
current_platform.is_cuda_alike(), reason="Skipping to avoid redundant model tests"
)
@multi_gpu_test(num_gpus=4)
def test_llama_lora_tp4(llama32_lora_files):
llm = vllm.LLM(