[CI] Run basic fullgraph correctness on one GPU (#51271)
Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
co-authored by
OpenAI Codex
parent
7b4ed49628
commit
566c80edf9
@@ -68,14 +68,12 @@ steps:
|
||||
- vllm/worker/worker_base.py
|
||||
- vllm/v1/engine/
|
||||
- vllm/v1/worker/
|
||||
- tests/compile/fullgraph/test_basic_correctness.py
|
||||
- tests/compile/test_wrapper.py
|
||||
- tests/entrypoints/llm/test_collective_rpc.py
|
||||
commands:
|
||||
# https://github.com/NVIDIA/nccl/issues/1838
|
||||
- export NCCL_CUMEM_HOST_ENABLE=0
|
||||
- pytest -v -s entrypoints/llm/test_collective_rpc.py
|
||||
- pytest -v -s ./compile/fullgraph/test_basic_correctness.py
|
||||
- pytest -v -s ./compile/test_wrapper.py
|
||||
|
||||
- label: Distributed Torchrun + Shutdown Tests (2 GPUs)
|
||||
@@ -163,13 +161,11 @@ steps:
|
||||
- vllm/distributed/
|
||||
- tests/distributed/test_pynccl
|
||||
- tests/distributed/test_events
|
||||
- tests/compile/fullgraph/test_basic_correctness.py
|
||||
- tests/distributed/test_symm_mem_allreduce.py
|
||||
- tests/distributed/test_multiproc_executor.py
|
||||
commands:
|
||||
# https://github.com/NVIDIA/nccl/issues/1838
|
||||
- export NCCL_CUMEM_HOST_ENABLE=0
|
||||
- pytest -v -s compile/fullgraph/test_basic_correctness.py
|
||||
- pytest -v -s distributed/test_pynccl.py
|
||||
- pytest -v -s distributed/test_events.py
|
||||
- pytest -v -s distributed/test_symm_mem_allreduce.py
|
||||
|
||||
@@ -27,15 +27,6 @@ class TestSetting:
|
||||
@pytest.mark.parametrize(
|
||||
"test_setting",
|
||||
[
|
||||
# basic llama model
|
||||
TestSetting(
|
||||
model="meta-llama/Llama-3.2-1B-Instruct",
|
||||
model_args=["--max-model-len", "2048"],
|
||||
pp_size=2,
|
||||
tp_size=2,
|
||||
attn_backend=ATTN_BACKEND,
|
||||
method="generate",
|
||||
),
|
||||
# MoE model
|
||||
TestSetting(
|
||||
model="ibm-granite/granite-3.0-1b-a400m-instruct",
|
||||
@@ -68,21 +59,12 @@ class TestSetting:
|
||||
def test_compile_correctness(
|
||||
test_setting: TestSetting,
|
||||
):
|
||||
# this test is run under multiple suits, with different GPUs.
|
||||
# make sure we only run the test with correct CUDA devices.
|
||||
# don't use "<", as it will duplicate the tests.
|
||||
model = test_setting.model
|
||||
model_args = test_setting.model_args
|
||||
pp_size = test_setting.pp_size
|
||||
tp_size = test_setting.tp_size
|
||||
attn_backend = test_setting.attn_backend
|
||||
method = test_setting.method
|
||||
if current_platform.device_count() < pp_size * tp_size:
|
||||
pytest.skip(
|
||||
f"Need at least {pp_size}*{tp_size} CUDA gpus but got "
|
||||
f"{current_platform.device_count()}"
|
||||
)
|
||||
|
||||
final_args = [
|
||||
*model_args,
|
||||
"-pp",
|
||||
|
||||
Reference in New Issue
Block a user