mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-19 13:52:07 +00:00
eecca27113
* feat: tokenize each request individually and increase warmup image size * feat: adjust rotary embed and avoid cuda graphs of size 2 and smaller * fix: address image resize and rebase changes * feat: update to run qwen2-vl tests * fix: tweak param types
39 lines
957 B
Python
39 lines
957 B
Python
import pytest
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def flash_qwen2_vl_handle(launcher):
|
|
with launcher(
|
|
"Qwen/Qwen2-VL-2B-Instruct",
|
|
max_input_length=40,
|
|
max_batch_prefill_tokens=50,
|
|
max_total_tokens=51,
|
|
) as handle:
|
|
yield handle
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
async def flash_qwen2(flash_qwen2_vl_handle):
|
|
await flash_qwen2_vl_handle.health(300)
|
|
return flash_qwen2_vl_handle.client
|
|
|
|
|
|
@pytest.mark.private
|
|
async def test_flash_qwen2_vl_simple(flash_qwen2, response_snapshot):
|
|
response = await flash_qwen2.chat(
|
|
max_tokens=20,
|
|
seed=42,
|
|
messages=[
|
|
{
|
|
"role": "user",
|
|
"content": [
|
|
{"type": "text", "text": "What is the color of the sky?"},
|
|
],
|
|
},
|
|
],
|
|
)
|
|
|
|
assert response.choices[0].message.content == "The correct answer is: blue"
|
|
|
|
assert response == response_snapshot
|