mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-25 20:12:07 +00:00
43 lines
1013 B
Python
43 lines
1013 B
Python
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)
|