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

Consolidate Llama model usage in tests (#13094)

Browse Source
This commit is contained in:
Harry Mellor
2025-02-14 06:18:03 +00:00
committed by GitHub
parent 40932d7a05
commit f2b20fe491
22 changed files with 45 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/basic_correctness/test_basic_correctness.py
View File

@@ -17,7 +17,7 @@ from ..utils import multi_gpu_test
MODELS = [
"google/gemma-2-2b-it",
"meta-llama/Llama-3.2-1B",
"meta-llama/Llama-3.2-1B-Instruct",
]
TARGET_TEST_SUITE = os.environ.get("TARGET_TEST_SUITE", "L4")
@@ -96,12 +96,12 @@ def test_models(
"test_suite", [
("facebook/opt-125m", "ray", "", "L4"),
("facebook/opt-125m", "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"),
("facebook/opt-125m", "ray", "", "A100"),
("facebook/opt-125m", "mp", "", "A100"),
("facebook/opt-125m", "mp", "FLASHINFER", "A100"),
("meta-llama/Meta-Llama-3-8B", "ray", "FLASHINFER", "A100"),
("meta-llama/Llama-3.2-1B-Instruct", "ray", "FLASHINFER", "A100"),
])
def test_models_distributed(
hf_runner,
@@ -116,7 +116,7 @@ def test_models_distributed(
if test_suite != TARGET_TEST_SUITE:
pytest.skip(f"Skip test for {test_suite}")
if model == "meta-llama/Llama-2-7b-hf" and distributed_executor_backend == "ray" and attention_backend == "" and test_suite == "L4": # noqa
if model == "meta-llama/Llama-3.2-1B-Instruct" and distributed_executor_backend == "ray" and attention_backend == "" and test_suite == "L4": # noqa
# test ray adag
os.environ['VLLM_USE_RAY_SPMD_WORKER'] = "1"
os.environ['VLLM_USE_RAY_COMPILED_DAG'] = "1"

6
tests/basic_correctness/test_chunked_prefill.py
View File

@@ -20,7 +20,7 @@ from ..utils import multi_gpu_test
MODELS = [
"facebook/opt-125m",
"meta-llama/Llama-3.2-1B",
"meta-llama/Llama-3.2-1B-Instruct",
]
@@ -92,7 +92,7 @@ def test_models_distributed(
) -> None:
override_backend_env_variable(monkeypatch, attention_backend)
if (model == "meta-llama/Llama-2-7b-hf"
if (model == "meta-llama/Llama-3.2-1B-Instruct"
and distributed_executor_backend == "ray"):
# test ray adag
os.environ['VLLM_USE_RAY_SPMD_WORKER'] = "1"
@@ -221,7 +221,7 @@ def test_with_prefix_caching(
Checks exact match decode with and without prefix caching
with chunked prefill enabled.
"""
model = "meta-llama/Llama-2-7b-chat-hf"
model = "meta-llama/Llama-3.2-1B-Instruct"
# The common prompt has 142 tokens with Llama-2 tokenizer.
common_prompt = "You are a helpful AI assistant " * 20
unique_prompts = [

2
tests/basic_correctness/test_cpu_offload.py
View File

@@ -4,5 +4,5 @@ from ..utils import compare_two_settings
def test_cpu_offload():
compare_two_settings("meta-llama/Llama-3.2-1B", [],
compare_two_settings("meta-llama/Llama-3.2-1B-Instruct", [],
["--cpu-offload-gb", "1"])

2
tests/basic_correctness/test_cumem.py
View File

@@ -118,7 +118,7 @@ def test_cumem_with_cudagraph():
@pytest.mark.parametrize(
"model",
[
"meta-llama/Llama-3.2-1B", # sleep mode with safetensors
"meta-llama/Llama-3.2-1B-Instruct", # sleep mode with safetensors
"facebook/opt-125m" # sleep mode with pytorch checkpoint
])
def test_end_to_end(model):