biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI] Replace large models with tiny alternatives in tests (#24057)

Browse Source 
Signed-off-by: Tahsin Tunan <tahsintunan@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Tahsin Tunan
2025-10-16 20:51:27 +06:00
committed by GitHub
parent 02d709a6f1
commit 43721bc67f
17 changed files with 118 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/v1/shutdown/test_startup_error.py
View File

@@ -16,7 +16,7 @@ from vllm.model_executor.models.llama import LlamaForCausalLM
from vllm.utils import cuda_device_count_stateless
from vllm.v1.engine.async_llm import AsyncLLM
MODELS = ["meta-llama/Llama-3.2-1B"]
MODELS = ["hmellor/tiny-random-LlamaForCausalLM"]
def evil_method(self, *args, **kwargs):
@@ -76,8 +76,10 @@ def test_llm_startup_error(
Test profiling (forward()) and load weights failures.
TODO(andy) - LLM without multiprocessing.
"""
if model != "meta-llama/Llama-3.2-1B":
pytest.skip(reason="Only test meta-llama/Llama-3.2-1B")
# Skip non-Llama models since we monkeypatch LlamaForCausalLM specifically.
# If MODELS list grows, each architecture needs its own test variant.
if model != "JackFram/llama-68m":
pytest.skip(reason="Only test JackFram/llama-68m")
if cuda_device_count_stateless() < tensor_parallel_size:
pytest.skip(reason="Not enough CUDA devices")