vllm/tests/models/test_mistral.py

38 lines
1.2 KiB
Python

"""Compare the outputs of HF and vLLM for Mistral models using greedy sampling.
Run `pytest tests/models/test_mistral.py --forked`.
"""
import pytest
MODELS = [
"mistralai/Mistral-7B-Instruct-v0.1",
]
@pytest.mark.parametrize("model", MODELS)
@pytest.mark.parametrize("dtype", ["bfloat16"])
@pytest.mark.parametrize("max_tokens", [128])
def test_models(
hf_runner,
vllm_runner,
example_long_prompts,
model: str,
dtype: str,
max_tokens: int,
) -> None:
hf_model = hf_runner(model, dtype=dtype)
hf_outputs = hf_model.generate_greedy(example_long_prompts, max_tokens)
del hf_model
vllm_model = vllm_runner(model, dtype=dtype)
vllm_outputs = vllm_model.generate_greedy(example_long_prompts, max_tokens)
del vllm_model
for i in range(len(example_long_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}")