42 lines
1.3 KiB
Python
42 lines
1.3 KiB
Python
|
"""Compare the outputs of HF and distributed vLLM when using greedy sampling.
|
||
|
|
|
||
|
|
Run `pytest tests/distributed/test_basic_distributed_correctness.py --forked`.
|
||
|
|
"""
|
||
|
|
import pytest
|
||
|
|
import torch
|
||
|
|
|
||
|
|
MODELS = [
|
||
|
|
"facebook/opt-125m",
|
||
|
|
"meta-llama/Llama-2-7b-hf",
|
||
|
|
]
|
||
|
|
|
||
|
|
|
||
|
|
@pytest.mark.skipif(torch.cuda.device_count() < 2,
|
||
|
|
reason="Need at least 2 GPUs to run the test.")
|
||
|
|
@pytest.mark.parametrize("model", MODELS)
|
||
|
|
@pytest.mark.parametrize("dtype", ["half"])
|
||
|
|
@pytest.mark.parametrize("max_tokens", [5])
|
||
|
|
def test_models(
|
||
|
|
hf_runner,
|
||
|
|
vllm_runner,
|
||
|
|
example_prompts,
|
||
|
|
model: str,
|
||
|
|
dtype: str,
|
||
|
|
max_tokens: int,
|
||
|
|
) -> None:
|
||
|
|
hf_model = hf_runner(model, dtype=dtype)
|
||
|
|
hf_outputs = hf_model.generate_greedy(example_prompts, max_tokens)
|
||
|
|
del hf_model
|
||
|
|
|
||
|
|
vllm_model = vllm_runner(model, dtype=dtype, tensor_parallel_size=2)
|
||
|
|
vllm_outputs = vllm_model.generate_greedy(example_prompts, max_tokens)
|
||
|
|
del vllm_model
|
||
|
|
|
||
|
|
for i in range(len(example_prompts)):
|
||
|
|
hf_output_ids, hf_output_str = hf_outputs[i]
|
||
|
|
vllm_output_ids, vllm_output_str = vllm_outputs[i]
|
||
|
|
assert hf_output_str == vllm_output_str, (
|
||
|
|
f"Test{i}:\nHF: {hf_output_str!r}\nvLLM: {vllm_output_str!r}")
|
||
|
|
assert hf_output_ids == vllm_output_ids, (
|
||
|
|
f"Test{i}:\nHF: {hf_output_ids}\nvLLM: {vllm_output_ids}")
|