Skip to content

[CI] Update Bagel Pixels #4081

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Gaohan123 merged 7 commits into from
Jun 4, 2026
Merged

[CI] Update Bagel Pixels #4081

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
e840a24
disable lance changes (passing bagel tests)
alex-jw-brooks Jun 2, 2026
0c43b71
update ref pixels, bring lance changes back
alex-jw-brooks Jun 2, 2026
2d011b0
cuda only for bagel tests
alex-jw-brooks Jun 3, 2026
f3f3058
re-enable back on mi325
alex-jw-brooks Jun 4, 2026
5b3a9ea
revert lance on device latent changes, use 14 steps for off by one fix
alex-jw-brooks Jun 4, 2026
8865f50
comment fix
alex-jw-brooks Jun 4, 2026
c09ddb7
Merge branch 'main' into enable_bagel_test
hsliuustc0106 Jun 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 3 additions & 4 deletions tests/distributed/omni_connectors/test_bagel_mooncake_connector.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,8 @@

# Reference pixel data extracted from the known-good output image
# Each entry contains (x, y) position and expected (R, G, B) values
# "Generated with seed=52, num_inference_steps=15,
# prompt='A futuristic city skyline at twilight, cyberpunk style'"
# "Generated with seed=52, num_inference_steps=14,
# prompt='A cute cat'"
REFERENCE_PIXELS = [
{"position": (100, 100), "rgb": (115, 113, 94)},
{"position": (400, 50), "rgb": (159, 160, 144)},
Expand Down Expand Up @@ -63,7 +63,7 @@ def _find_free_port() -> int:
return port


def _configure_sampling_params(omni: Omni, num_inference_steps: int = 15) -> list:
def _configure_sampling_params(omni: Omni, num_inference_steps: int = 14) -> list:
"""Configure sampling parameters for Bagel text2img generation.

Args:
Expand Down Expand Up @@ -269,7 +269,6 @@ def _load_mooncake_config(host: str, rpc_port: int, http_port: int) -> str:
@pytest.mark.core_model
@pytest.mark.advanced_model
@pytest.mark.diffusion
@pytest.mark.skip(reason="Skip failed CI issue 3977: https://github.com/vllm-project/vllm-omni/issues/3977")
@hardware_test(res={"cuda": "H100"}, num_cards=1)
def test_bagel_text2img_mooncake_connector(run_level):
"""Test Bagel text2img with Mooncake connector for inter-stage communication."""
Expand Down
37 changes: 4 additions & 33 deletions tests/distributed/omni_connectors/test_bagel_shared_memory_connector.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,14 +24,13 @@
from tests.helpers.runtime import OmniRunner
from tests.helpers.stage_config import get_deploy_config_path, modify_stage_config
from vllm_omni.entrypoints.omni import Omni
from vllm_omni.platforms import current_omni_platform

pytestmark = [pytest.mark.usefixtures("clean_gpu_memory_between_tests")]

BAGEL_CI_DEPLOY = get_deploy_config_path("ci/bagel.yaml")

# Reference pixel data extracted from the known-good output image
# Generated with seed=52, num_inference_steps=15,
# Generated with seed=52, num_inference_steps=14,
# prompt='Change the grass color to red',
# input image: 2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg
REFERENCE_PIXELS = [
Expand All @@ -47,23 +46,10 @@
{"position": (256, 256), "rgb": (181, 201, 221)},
]

if current_omni_platform.is_rocm():
REFERENCE_PIXELS = [

@alex-jw-brooks alex-jw-brooks Jun 2, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure why rocm pixels are redefined here, but since they were exactly the same, I deleted them. Unfortunately don't have a rocm device to test the new values on 😅

{"position": (100, 100), "rgb": (156, 172, 217)},
{"position": (400, 50), "rgb": (105, 144, 217)},
{"position": (700, 100), "rgb": (118, 159, 232)},
{"position": (150, 400), "rgb": (180, 22, 52)},
{"position": (512, 336), "rgb": (221, 211, 194)},
{"position": (700, 400), "rgb": (192, 10, 46)},
{"position": (100, 600), "rgb": (102, 12, 22)},
{"position": (400, 600), "rgb": (161, 28, 47)},
{"position": (700, 600), "rgb": (100, 87, 94)},
{"position": (256, 256), "rgb": (181, 201, 221)},
]

# text2img reference pixels (aligned with offline `bagel/end2end.py` text2img, 15 steps)
# "Generated with seed=52, num_inference_steps=15,
# prompt='A futuristic city skyline at twilight, cyberpunk style'"
# "Generated with seed=52, num_inference_steps=14,
# prompt='A cute cat'"
TEXT2IMG_REFERENCE_PIXELS = [
{"position": (100, 100), "rgb": (115, 113, 94)},
{"position": (400, 50), "rgb": (159, 160, 144)},
Expand All @@ -77,19 +63,6 @@
{"position": (256, 256), "rgb": (167, 156, 150)},
]

if current_omni_platform.is_rocm():
TEXT2IMG_REFERENCE_PIXELS = [
{"position": (100, 100), "rgb": (115, 113, 94)},
{"position": (400, 50), "rgb": (159, 160, 144)},
{"position": (700, 100), "rgb": (164, 151, 123)},
{"position": (150, 400), "rgb": (120, 121, 107)},
{"position": (512, 512), "rgb": (165, 133, 127)},
{"position": (700, 400), "rgb": (217, 130, 66)},
{"position": (100, 700), "rgb": (191, 168, 152)},
{"position": (400, 700), "rgb": (130, 96, 77)},
{"position": (700, 700), "rgb": (247, 203, 140)},
{"position": (256, 256), "rgb": (167, 156, 150)},
]

PIXEL_TOLERANCE = 10
TEXT2IMG_PIXEL_TOLERANCE = 5
Expand All @@ -115,7 +88,7 @@ def _find_free_port() -> int:
return port


def _configure_sampling_params(omni: Omni, num_inference_steps: int = 15) -> list:
def _configure_sampling_params(omni: Omni, num_inference_steps: int = 14) -> list:
"""Configure sampling parameters for Bagel img2img generation.

Args:
Expand Down Expand Up @@ -258,7 +231,6 @@ def _resolve_deploy_config(config_path: str, run_level: str) -> str:
@pytest.mark.core_model
@pytest.mark.advanced_model
@pytest.mark.diffusion
@pytest.mark.skip(reason="Skip failed CI issue 3977: https://github.com/vllm-project/vllm-omni/issues/3977")
@hardware_test(res={"cuda": "H100", "rocm": "MI325"})
def test_bagel_img2img_shared_memory_connector(run_level):
"""Test Bagel img2img with shared memory connector."""
Expand All @@ -276,7 +248,6 @@ def test_bagel_img2img_shared_memory_connector(run_level):
@pytest.mark.core_model
@pytest.mark.advanced_model
@pytest.mark.diffusion
@pytest.mark.skip(reason="Skip failed CI issue 3977: https://github.com/vllm-project/vllm-omni/issues/3977")
@hardware_test(res={"cuda": "H100", "rocm": "MI325"})
def test_bagel_text2img_shared_memory_connector(run_level):
"""Test Bagel text2img with shared memory connector."""
Expand Down
4 changes: 3 additions & 1 deletion vllm_omni/diffusion/models/bagel/pipeline_bagel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -777,7 +777,9 @@ def vae_transforms(img):
if torch.is_tensor(v):
generation_input[k] = v.to(self.device)

self._regen_init_noise_on_device(generation_input, req.sampling_params.seed)
# NOTE: For now we disable device specific noise regeneration so that e2e tests can run
# on both CUDA and ROCm. Context: https://github.com/vllm-project/vllm-omni/pull/4081
# self._regen_init_noise_on_device(generation_input, req.sampling_params.seed)

# text cfg
generation_input_cfg_text = self.bagel.prepare_vae_latent_cfg(
Expand Down
Loading