test_flash_neox.py 1.13 KB
Newer Older
1
2
import pytest

3
4
5
6
7

@pytest.fixture(scope="module")
def flash_neox_handle(launcher):
    with launcher("OpenAssistant/oasst-sft-1-pythia-12b", num_shard=2) as handle:
        yield handle
8
9
10


@pytest.fixture(scope="module")
11
12
13
async def flash_neox(flash_neox_handle):
    await flash_neox_handle.health(240)
    return flash_neox_handle.client
14
15
16


@pytest.mark.asyncio
17
async def test_flash_neox(flash_neox, response_snapshot):
18
19
20
21
22
23
    response = await flash_neox.generate(
        "<|prompter|>What is a meme, and what's the history behind this word?<|endoftext|><|assistant|>",
        max_new_tokens=10,
    )

    assert response.details.generated_tokens == 10
24
    assert response == response_snapshot
25
26
27


@pytest.mark.asyncio
28
async def test_flash_neox_load(flash_neox, generate_load, response_snapshot):
29
30
31
32
33
34
35
36
    responses = await generate_load(
        flash_neox,
        "<|prompter|>What is a meme, and what's the history behind this word?<|endoftext|><|assistant|>",
        max_new_tokens=10,
        n=4,
    )

    assert len(responses) == 4
37
    assert all([r.generated_text == responses[0].generated_text for r in responses])
38

39
    assert responses == response_snapshot