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

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

76
tests/lora/test_transformers_model.py
View File

@@ -24,20 +24,18 @@ def do_sample(llm: vllm.LLM, lora_path: str, lora_id: int) -> list[str]:
prompts = [
PROMPT_TEMPLATE.format(query="How many singers do we have?"),
PROMPT_TEMPLATE.format(
query=
"What is the average, minimum, and maximum age of all singers from France?" # noqa: E501
query="What is the average, minimum, and maximum age of all singers from France?" # noqa: E501
),
PROMPT_TEMPLATE.format(
query=
"What are all distinct countries where singers above age 20 are from?" # noqa: E501
query="What are all distinct countries where singers above age 20 are from?" # noqa: E501
),
]
sampling_params = vllm.SamplingParams(temperature=0, max_tokens=32)
outputs = llm.generate(
prompts,
sampling_params,
lora_request=LoRARequest(str(lora_id), lora_id, lora_path)
if lora_id else None)
lora_request=LoRARequest(str(lora_id), lora_id, lora_path) if lora_id else None,
)
# Print the outputs.
generated_texts: list[str] = []
for output in outputs:
@@ -49,13 +47,15 @@ def do_sample(llm: vllm.LLM, lora_path: str, lora_id: int) -> list[str]:
def test_ilama_lora(ilama_lora_files):
llm = vllm.LLM(MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
trust_remote_code=True,
enable_chunked_prefill=True)
llm = vllm.LLM(
MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
trust_remote_code=True,
enable_chunked_prefill=True,
)
output1 = do_sample(llm, ilama_lora_files, lora_id=1)
for i in range(len(EXPECTED_LORA_OUTPUT)):
@@ -65,20 +65,23 @@ def test_ilama_lora(ilama_lora_files):
assert output2[i] == EXPECTED_LORA_OUTPUT[i]
@pytest.mark.skipif(current_platform.is_cuda_alike(),
reason="Skipping to avoid redundant model tests")
@pytest.mark.skipif(
current_platform.is_cuda_alike(), reason="Skipping to avoid redundant model tests"
)
@multi_gpu_test(num_gpus=4)
@create_new_process_for_each_test()
def test_ilama_lora_tp4(ilama_lora_files):
llm = vllm.LLM(MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
tensor_parallel_size=4,
trust_remote_code=True,
fully_sharded_loras=False,
enable_chunked_prefill=True)
llm = vllm.LLM(
MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
tensor_parallel_size=4,
trust_remote_code=True,
fully_sharded_loras=False,
enable_chunked_prefill=True,
)
output1 = do_sample(llm, ilama_lora_files, lora_id=1)
for i in range(len(EXPECTED_LORA_OUTPUT)):
@@ -88,20 +91,23 @@ def test_ilama_lora_tp4(ilama_lora_files):
assert output2[i] == EXPECTED_LORA_OUTPUT[i]
@pytest.mark.skipif(current_platform.is_cuda_alike(),
reason="Skipping to avoid redundant model tests")
@pytest.mark.skipif(
current_platform.is_cuda_alike(), reason="Skipping to avoid redundant model tests"
)
@multi_gpu_test(num_gpus=4)
@create_new_process_for_each_test()
def test_ilama_lora_tp4_fully_sharded_loras(ilama_lora_files):
llm = vllm.LLM(MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
tensor_parallel_size=4,
trust_remote_code=True,
fully_sharded_loras=True,
enable_chunked_prefill=True)
llm = vllm.LLM(
MODEL_PATH,
max_model_len=1024,
enable_lora=True,
max_loras=4,
max_lora_rank=16,
tensor_parallel_size=4,
trust_remote_code=True,
fully_sharded_loras=True,
enable_chunked_prefill=True,
)
output1 = do_sample(llm, ilama_lora_files, lora_id=1)
for i in range(len(EXPECTED_LORA_OUTPUT)):
assert output1[i] == EXPECTED_LORA_OUTPUT[i]