[diffusion] platform: support WAN/FLUX/Qwen-Image/Qwen-Image-edit on Ascend (#13662)
Co-authored-by: dhx98 <haox.dai@gmail.com> Co-authored-by: DHX98 <haoxiand@andrew.cmu.edu> Co-authored-by: ronnie_zheng <zl19940307@163.com> Co-authored-by: DHX98 <DHX98@noreply.gitcode.com> Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
This commit is contained in:
Makcum888e
co-authored by
dhx98
DHX98
ronnie_zheng
DHX98
Yuhao Yang
dhx98
DHX98
ronnie_zheng
DHX98
Yuhao Yang
parent
7b83659310
commit
00248d85c7
@@ -38,6 +38,15 @@ SUITES = {
|
||||
],
|
||||
}
|
||||
|
||||
suites_ascend = {
|
||||
"1-npu": [
|
||||
"ascend/test_server_1_npu.py",
|
||||
# add new 1-npu test files here
|
||||
]
|
||||
}
|
||||
|
||||
SUITES.update(suites_ascend)
|
||||
|
||||
|
||||
def parse_args():
|
||||
parser = argparse.ArgumentParser(description="Run multimodal_gen test suite")
|
||||
|
||||
@@ -0,0 +1,76 @@
|
||||
{
|
||||
"metadata": {
|
||||
"model": "Diffusion Server",
|
||||
"hardware": "CI A2 64GB pool",
|
||||
"description": "Reference numbers captured from the CI diffusion server baseline run"
|
||||
},
|
||||
"scenarios": {
|
||||
"wan2_1_t2v_1.3b_1_npu": {
|
||||
"stages_ms": {
|
||||
"InputValidationStage": 0.1,
|
||||
"TextEncodingStage": 1609.27,
|
||||
"ConditioningStage": 0.02,
|
||||
"TimestepPreparationStage": 3.46,
|
||||
"LatentPreparationStage": 0.39,
|
||||
"DenoisingStage": 26324.0,
|
||||
"DecodingStage": 817.68,
|
||||
"per_frame_generation": null
|
||||
},
|
||||
"denoise_step_ms": {
|
||||
"0": 195.27,
|
||||
"1": 329.05,
|
||||
"2": 545.43,
|
||||
"3": 541.3,
|
||||
"4": 537.07,
|
||||
"5": 537.21,
|
||||
"6": 537.19,
|
||||
"7": 537.19,
|
||||
"8": 537.27,
|
||||
"9": 537.05,
|
||||
"10": 537.02,
|
||||
"11": 537.11,
|
||||
"12": 537.42,
|
||||
"13": 537.2,
|
||||
"14": 537.16,
|
||||
"15": 537.11,
|
||||
"16": 537.14,
|
||||
"17": 537.19,
|
||||
"18": 537.1,
|
||||
"19": 537.0,
|
||||
"20": 537.26,
|
||||
"21": 537.18,
|
||||
"22": 537.16,
|
||||
"23": 537.24,
|
||||
"24": 537.15,
|
||||
"25": 537.14,
|
||||
"26": 536.99,
|
||||
"27": 537.19,
|
||||
"28": 537.22,
|
||||
"29": 537.23,
|
||||
"30": 537.06,
|
||||
"31": 537.06,
|
||||
"32": 537.18,
|
||||
"33": 537.07,
|
||||
"34": 537.19,
|
||||
"35": 537.28,
|
||||
"36": 537.17,
|
||||
"37": 537.38,
|
||||
"38": 537.31,
|
||||
"39": 537.25,
|
||||
"40": 537.28,
|
||||
"41": 537.26,
|
||||
"42": 537.1,
|
||||
"43": 537.19,
|
||||
"44": 537.19,
|
||||
"45": 537.31,
|
||||
"46": 537.19,
|
||||
"47": 537.16,
|
||||
"48": 537.23,
|
||||
"49": 532.91
|
||||
},
|
||||
"expected_e2e_ms": 28769.9,
|
||||
"expected_avg_denoise_ms": 526.34,
|
||||
"expected_median_denoise_ms": 537.19
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,29 @@
|
||||
"""
|
||||
Config-driven diffusion performance test with pytest parametrization.
|
||||
|
||||
|
||||
If the actual run is significantly better than the baseline, the improved cases with their updated baseline will be printed
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import pytest
|
||||
|
||||
from sglang.multimodal_gen.runtime.utils.logging_utils import init_logger
|
||||
from sglang.multimodal_gen.test.server.ascend.testcase_configs_npu import ONE_NPU_CASES
|
||||
from sglang.multimodal_gen.test.server.test_server_common import ( # noqa: F401
|
||||
DiffusionServerBase,
|
||||
diffusion_server,
|
||||
)
|
||||
from sglang.multimodal_gen.test.server.testcase_configs import DiffusionTestCase
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class TestDiffusionServerOneNpu(DiffusionServerBase):
|
||||
"""Performance tests for 1-NPU diffusion cases."""
|
||||
|
||||
@pytest.fixture(params=ONE_NPU_CASES, ids=lambda c: c.id)
|
||||
def case(self, request) -> DiffusionTestCase:
|
||||
"""Provide a DiffusionTestCase for each 1-NPU test."""
|
||||
return request.param
|
||||
@@ -0,0 +1,22 @@
|
||||
from sglang.multimodal_gen.test.server.testcase_configs import (
|
||||
T2V_PROMPT,
|
||||
DiffusionSamplingParams,
|
||||
DiffusionServerArgs,
|
||||
DiffusionTestCase,
|
||||
)
|
||||
|
||||
ONE_NPU_CASES: list[DiffusionTestCase] = [
|
||||
# === Text to Video (T2V) ===
|
||||
DiffusionTestCase(
|
||||
"wan2_1_t2v_1.3b_1_npu",
|
||||
DiffusionServerArgs(
|
||||
model_path="/root/.cache/modelscope/hub/models/Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
|
||||
modality="video",
|
||||
warmup=0,
|
||||
custom_validator="video",
|
||||
),
|
||||
DiffusionSamplingParams(
|
||||
prompt=T2V_PROMPT,
|
||||
),
|
||||
),
|
||||
]
|
||||
@@ -132,6 +132,24 @@ class BaselineConfig:
|
||||
),
|
||||
)
|
||||
|
||||
def update(self, path: Path):
|
||||
"""Load baseline configuration from JSON file."""
|
||||
with path.open("r", encoding="utf-8") as fh:
|
||||
data = json.load(fh)
|
||||
|
||||
scenarios_new = {}
|
||||
for name, cfg in data["scenarios"].items():
|
||||
scenarios_new[name] = ScenarioConfig(
|
||||
stages_ms=cfg["stages_ms"],
|
||||
denoise_step_ms={int(k): v for k, v in cfg["denoise_step_ms"].items()},
|
||||
expected_e2e_ms=float(cfg["expected_e2e_ms"]),
|
||||
expected_avg_denoise_ms=float(cfg["expected_avg_denoise_ms"]),
|
||||
expected_median_denoise_ms=float(cfg["expected_median_denoise_ms"]),
|
||||
)
|
||||
|
||||
self.scenarios.update(scenarios_new)
|
||||
return self
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class DiffusionServerArgs:
|
||||
@@ -729,4 +747,6 @@ TWO_GPU_CASES_B = [
|
||||
]
|
||||
|
||||
# Load global configuration
|
||||
BASELINE_CONFIG = BaselineConfig.load(Path(__file__).with_name("perf_baselines.json"))
|
||||
BASELINE_CONFIG = BaselineConfig.load(
|
||||
Path(__file__).with_name("perf_baselines.json")
|
||||
).update(Path(__file__).parent / "ascend" / "perf_baselines_npu.json")
|
||||
|
||||
Reference in New Issue
Block a user