hf_text-generation-inference/integration-tests/models/test_bloom_560m_sharded.py

43 lines
1013 B
Python
Raw Normal View History

2023-05-15 15:36:30 -06:00
import pytest
from utils import health_check
@pytest.fixture(scope="module")
def bloom_560m_sharded(launcher):
with launcher("bigscience/bloom-560m", num_shard=2) as client:
yield client
@pytest.mark.asyncio
async def test_bloom_560m_sharded(bloom_560m_sharded, snapshot_test):
await health_check(bloom_560m_sharded, 60)
response = await bloom_560m_sharded.generate(
"Pour déguster un ortolan, il faut tout d'abord",
max_new_tokens=10,
top_p=0.9,
seed=0,
)
assert response.details.generated_tokens == 10
assert snapshot_test(response)
@pytest.mark.asyncio
async def test_bloom_560m_sharded_load(
bloom_560m_sharded, generate_load, snapshot_test
):
await health_check(bloom_560m_sharded, 60)
responses = await generate_load(
bloom_560m_sharded,
"Pour déguster un ortolan, il faut tout d'abord",
max_new_tokens=10,
n=4,
)
assert len(responses) == 4
assert snapshot_test(responses)