[Rebase] Align vLLM-Omni with vLLM v0.31.0 (#8459)

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
This commit is contained in:
Zhou Taichang authored and GitHub committed 2026-10-05 14:11:01 +08:00
1 parent 1f47fa2f29
commit 0b0d2d7009
141 files changed
+2885 -721

No files matched your search

@@ -3,6 +3,7 @@
import contextlib
from types import SimpleNamespace
from unittest.mock import Mock
import pytest
import torch
@@ -22,6 +23,23 @@ class _StopExecutionError(Exception):
"""Halts execute_model at the model boundary."""
@pytest.mark.parametrize("randomize_inputs", [False, True])
def test_generation_profile_accepts_upstream_randomize_inputs(randomize_inputs):
"""Release workers can pass the profiling flag to codec-stage runners."""
runner = object.__new__(GPUGenerationModelRunner)
runner.supports_mm_inputs = False
runner.max_num_tokens = 128
runner.encoder_cache = {"temporary": object()}
runner._dummy_run = Mock(return_value=(torch.zeros(1), None))
runner._sync_device = Mock()
runner.profile_run(randomize_inputs=randomize_inputs)
runner._dummy_run.assert_called_once_with(128, is_profile=True, randomize_inputs=randomize_inputs)
runner._sync_device.assert_called_once_with()
assert runner.encoder_cache == {}
class _DummyInputBatch:
def __init__(self, num_reqs: int = 1):
self.req_ids = [f"req-{i + 1}" for i in range(num_reqs)]
@@ -15,8 +15,8 @@ pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
def test_dispatch_eager_and_manager_paths(monkeypatch, dp_size):
runner = object.__new__(OmniGPUModelRunner)
runner.cudagraph_manager = None
runner.dp_size = dp_size
runner.dp_rank = 1
runner.parallel_config = SimpleNamespace(data_parallel_size=dp_size, data_parallel_rank=1)
runner.dp_size, runner.dp_rank = dp_size, 1
import vllm.v1.worker.gpu.dp_utils as dp_utils
@@ -52,6 +52,7 @@ def test_dispatch_eager_and_manager_paths(monkeypatch, dp_size):
return expected
runner.cudagraph_manager = SimpleNamespace(dispatch=dispatch)
runner.parallel_config.data_parallel_size = 1
runner.dp_size = 1
batch_desc, _ = runner._dispatch_batch_descriptor(
num_reqs=1, num_toks=8, uniform_tok_count=8, num_active_loras=0, use_eager=False, max_query_len=8