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

[ci] Use env var to control whether to use S3 bucket in CI (#13634)

Browse Source
This commit is contained in:
Kevin H. Luu
2025-02-22 19:19:45 -08:00
committed by GitHub
parent 322d2a27d6
commit 2c5e637b57
30 changed files with 222 additions and 231 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
tests/basic_correctness/test_basic_correctness.py
View File

@@ -9,7 +9,6 @@ import weakref
import pytest
from vllm import LLM
from vllm.config import LoadFormat
from vllm.platforms import current_platform
from ..conftest import VllmRunner
@@ -34,7 +33,7 @@ def v1(run_with_both_engines):
def test_vllm_gc_ed():
"""Verify vllm instance is GC'ed when it is deleted"""
llm = LLM("distilbert/distilgpt2", load_format=LoadFormat.RUNAI_STREAMER)
llm = LLM("distilbert/distilgpt2")
weak_llm = weakref.ref(llm)
del llm
# If there's any circular reference to vllm, this fails
@@ -43,10 +42,10 @@ def test_vllm_gc_ed():
@pytest.mark.parametrize("model", MODELS)
@pytest.mark.parametrize("backend", ["FLASH_ATTN", "XFORMERS", "FLASHINFER"])
@pytest.mark.parametrize("backend", ["FLASH_ATTN"])
@pytest.mark.parametrize("dtype", ["half"])
@pytest.mark.parametrize("max_tokens", [5])
@pytest.mark.parametrize("enforce_eager", [False, True])
@pytest.mark.parametrize("enforce_eager", [False])
def test_models(
hf_runner,
model: str,
@@ -97,8 +96,8 @@ def test_models(
"test_suite", [
("distilbert/distilgpt2", "ray", "", "L4"),
("distilbert/distilgpt2", "mp", "", "L4"),
("meta-llama/Llama-2-7b-hf", "ray", "", "L4"),
("meta-llama/Llama-2-7b-hf", "mp", "", "L4"),
("meta-llama/Llama-3.2-1B-Instruct", "ray", "", "L4"),
("meta-llama/Llama-3.2-1B-Instruct", "mp", "", "L4"),
("distilbert/distilgpt2", "ray", "", "A100"),
("distilbert/distilgpt2", "mp", "", "A100"),
("distilbert/distilgpt2", "mp", "FLASHINFER", "A100"),

9
tests/basic_correctness/test_cumem.py
View File

@@ -4,11 +4,9 @@ import pytest
import torch
from vllm import LLM, SamplingParams
from vllm.config import LoadFormat
from vllm.device_allocator.cumem import CuMemAllocator
from vllm.utils import GiB_bytes
from ..conftest import MODEL_WEIGHTS_S3_BUCKET
from ..utils import fork_new_process_for_each_test
@@ -121,7 +119,7 @@ def test_cumem_with_cudagraph():
"model, use_v1",
[
# sleep mode with safetensors
(f"{MODEL_WEIGHTS_S3_BUCKET}/meta-llama/Llama-3.2-1B", True),
("meta-llama/Llama-3.2-1B", True),
# sleep mode with pytorch checkpoint
("facebook/opt-125m", False),
])
@@ -130,10 +128,7 @@ def test_end_to_end(model: str, use_v1: bool):
os.environ["VLLM_USE_V1"] = "1" if use_v1 else "0"
free, total = torch.cuda.mem_get_info()
used_bytes_baseline = total - free # in case other process is running
load_format = LoadFormat.AUTO
if "Llama" in model:
load_format = LoadFormat.RUNAI_STREAMER
llm = LLM(model, load_format=load_format, enable_sleep_mode=True)
llm = LLM(model, enable_sleep_mode=True)
prompt = "How are you?"
sampling_params = SamplingParams(temperature=0, max_tokens=10)
output = llm.generate(prompt, sampling_params)