[BugFix] Surface diffusion metrics in chat completions; sanitize Bagel img2img mm kwargs (#2932)
Signed-off-by: NumberWan <wantszkin2003@gmail.com>
This commit is contained in:
1 parent
d8cc7a0409
commit
ead87aa521
6 files changed
+173
-4
No files matched your search
@@ -0,0 +1,107 @@
|
||||
import pytest
|
||||
|
||||
from benchmarks.diffusion.backends import RequestFuncInput, async_request_chat_completions
|
||||
|
||||
|
||||
class _MockResponse:
|
||||
def __init__(self, payload: dict, status: int = 200):
|
||||
self._payload = payload
|
||||
self.status = status
|
||||
|
||||
async def __aenter__(self):
|
||||
return self
|
||||
|
||||
async def __aexit__(self, exc_type, exc, tb):
|
||||
return False
|
||||
|
||||
async def json(self):
|
||||
return self._payload
|
||||
|
||||
async def text(self):
|
||||
return str(self._payload)
|
||||
|
||||
|
||||
class _MockSession:
|
||||
def __init__(self, payload: dict):
|
||||
self._payload = payload
|
||||
|
||||
def post(self, *args, **kwargs):
|
||||
return _MockResponse(self._payload)
|
||||
|
||||
|
||||
@pytest.mark.core_model
|
||||
@pytest.mark.benchmark
|
||||
@pytest.mark.cpu
|
||||
@pytest.mark.asyncio
|
||||
async def test_chat_completions_metrics_fallback_to_top_level():
|
||||
payload = {
|
||||
"choices": [
|
||||
{
|
||||
"message": {
|
||||
"content": [
|
||||
{
|
||||
"type": "image_url",
|
||||
"image_url": {"url": "data:image/png;base64,abc"},
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
],
|
||||
"metrics": {
|
||||
"stage_durations": {"diffusion": 1.25},
|
||||
"peak_memory_mb": 4096.0,
|
||||
},
|
||||
}
|
||||
|
||||
output = await async_request_chat_completions(
|
||||
RequestFuncInput(
|
||||
prompt="draw a cat",
|
||||
api_url="http://test.local/v1/chat/completions",
|
||||
model="ByteDance-Seed/BAGEL-7B-MoT",
|
||||
),
|
||||
session=_MockSession(payload),
|
||||
)
|
||||
|
||||
assert output.success is True
|
||||
assert output.stage_durations == {"diffusion": 1.25}
|
||||
assert output.peak_memory_mb == 4096.0
|
||||
|
||||
|
||||
@pytest.mark.core_model
|
||||
@pytest.mark.benchmark
|
||||
@pytest.mark.cpu
|
||||
@pytest.mark.asyncio
|
||||
async def test_chat_completions_metrics_message_level_takes_precedence():
|
||||
payload = {
|
||||
"choices": [
|
||||
{
|
||||
"message": {
|
||||
"content": [
|
||||
{
|
||||
"type": "image_url",
|
||||
"image_url": {"url": "data:image/png;base64,abc"},
|
||||
"stage_durations": {"message_stage": 0.7},
|
||||
"peak_memory_mb": 1234.0,
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
],
|
||||
"metrics": {
|
||||
"stage_durations": {"top_level_stage": 9.9},
|
||||
"peak_memory_mb": 9999.0,
|
||||
},
|
||||
}
|
||||
|
||||
output = await async_request_chat_completions(
|
||||
RequestFuncInput(
|
||||
prompt="draw a dog",
|
||||
api_url="http://test.local/v1/chat/completions",
|
||||
model="ByteDance-Seed/BAGEL-7B-MoT",
|
||||
),
|
||||
session=_MockSession(payload),
|
||||
)
|
||||
|
||||
assert output.success is True
|
||||
assert output.stage_durations == {"message_stage": 0.7}
|
||||
assert output.peak_memory_mb == 1234.0
|
||||
@@ -107,7 +107,7 @@ def test_bagel_text2img_online(omni_server, openai_client) -> None:
|
||||
@hardware_test(res={"cuda": "H100", "rocm": "MI325"})
|
||||
@pytest.mark.parametrize("omni_server", test_params, indirect=True)
|
||||
def test_bagel_img2img_online(omni_server, openai_client) -> None:
|
||||
"""Test Bagel img2img via OpenAI-compatible chat completions API."""
|
||||
"""Test Bagel img2img with explicit height/width in chat completions API."""
|
||||
input_image = ImageAsset("2560px-Gfp-wisconsin-madison-the-nature-boardwalk").pil_image.convert("RGB")
|
||||
buffer = BytesIO()
|
||||
input_image.save(buffer, format="JPEG")
|
||||
@@ -118,6 +118,8 @@ def test_bagel_img2img_online(omni_server, openai_client) -> None:
|
||||
"messages": _build_img2img_messages(IMG2IMG_PROMPT, image_b64),
|
||||
"modalities": ["image"],
|
||||
"extra_body": {
|
||||
"height": 512,
|
||||
"width": 512,
|
||||
"num_inference_steps": 2,
|
||||
"guidance_scale": 0.0,
|
||||
"seed": 42,
|
||||
|
||||
@@ -1,6 +1,8 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
"""Unit tests for OmniChatCompletionResponse/StreamResponse metrics field."""
|
||||
|
||||
from types import SimpleNamespace
|
||||
|
||||
import pytest
|
||||
|
||||
pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
|
||||
@@ -50,3 +52,36 @@ def test_omni_chat_completion_stream_response_metrics():
|
||||
)
|
||||
assert response.modality == "audio"
|
||||
assert response.metrics == {"stage_latency": 0.5}
|
||||
|
||||
|
||||
def test_create_image_choice_exposes_diffusion_metrics():
|
||||
"""Ensure image chat content exposes profiler metrics for clients."""
|
||||
from PIL import Image
|
||||
|
||||
from vllm_omni.entrypoints.openai.serving_chat import OmniOpenAIServingChat
|
||||
|
||||
stage_durations = {"prefill": 0.12, "diffusion": 1.23}
|
||||
peak_memory_mb = 3210.5
|
||||
omni_outputs = SimpleNamespace(
|
||||
request_output=None,
|
||||
stage_durations=stage_durations,
|
||||
peak_memory_mb=peak_memory_mb,
|
||||
images=[Image.new("RGB", (2, 2), color=(255, 0, 0))],
|
||||
)
|
||||
|
||||
choices = OmniOpenAIServingChat._create_image_choice( # type: ignore[misc]
|
||||
None,
|
||||
omni_outputs=omni_outputs,
|
||||
role="assistant",
|
||||
request=SimpleNamespace(return_token_ids=False),
|
||||
)
|
||||
|
||||
assert len(choices) == 1
|
||||
content = choices[0].message.content
|
||||
assert isinstance(content, list)
|
||||
assert len(content) == 1
|
||||
first_item = content[0]
|
||||
assert first_item["type"] == "image_url"
|
||||
assert first_item["image_url"]["url"].startswith("data:image/png;base64,")
|
||||
assert first_item["stage_durations"] == stage_durations
|
||||
assert first_item["peak_memory_mb"] == peak_memory_mb
|
||||
Reference in new issue
Block a user