[Test] Unify L2/L3 test layout, Buildkite steps, and test helpers (#2556)
Signed-off-by: wangyu <410167048@qq.com> Signed-off-by: Huang, Zeyu <11222265+fhfuih@users.noreply.github.com> Signed-off-by: wangyu <53896905+yenuo26@users.noreply.github.com> Co-authored-by: Huang, Zeyu <11222265+fhfuih@users.noreply.github.com>
This commit is contained in:
1 parent
b85833eeec
commit
6f2ad7b403
134 files changed
+4904
-4185
No files matched your search
@@ -1,8 +1,12 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import functools
|
||||
from threading import Lock
|
||||
from types import SimpleNamespace
|
||||
from typing import TYPE_CHECKING
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
@@ -11,11 +15,20 @@ from vllm.v1.outputs import SamplerOutput
|
||||
from vllm.v1.sample.logits_processor.state import LogitsProcessors
|
||||
from vllm.v1.sample.metadata import SamplingMetadata
|
||||
|
||||
from vllm_omni.model_executor.models.cosyvoice3.cosyvoice3 import CosyVoice3Model
|
||||
from vllm_omni.worker.gpu_ar_model_runner import GPUARModelRunner
|
||||
|
||||
pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from vllm_omni.model_executor.models.cosyvoice3.cosyvoice3 import CosyVoice3Model
|
||||
|
||||
|
||||
@functools.lru_cache(maxsize=1)
|
||||
def _cosyvoice3_model_and_runner():
|
||||
"""Defer heavy Omni/vLLM imports until a test runs (avoids duplicate CustomOp init)."""
|
||||
from vllm_omni.model_executor.models.cosyvoice3.cosyvoice3 import CosyVoice3Model
|
||||
from vllm_omni.worker.gpu_ar_model_runner import GPUARModelRunner
|
||||
|
||||
return CosyVoice3Model, GPUARModelRunner
|
||||
|
||||
|
||||
class _DummyCode2Wav:
|
||||
def __init__(
|
||||
@@ -59,6 +72,7 @@ def _make_code2wav_model(
|
||||
num_samples: int = 32,
|
||||
outputs: list[tuple[torch.Tensor, dict[str, object] | None]] | None = None,
|
||||
) -> CosyVoice3Model:
|
||||
CosyVoice3Model, _ = _cosyvoice3_model_and_runner()
|
||||
model = object.__new__(CosyVoice3Model)
|
||||
nn.Module.__init__(model)
|
||||
model.model_stage = "cosyvoice3_code2wav"
|
||||
@@ -79,6 +93,7 @@ def _make_code2wav_model(
|
||||
|
||||
|
||||
def _make_talker_model() -> CosyVoice3Model:
|
||||
CosyVoice3Model, _ = _cosyvoice3_model_and_runner()
|
||||
model = object.__new__(CosyVoice3Model)
|
||||
nn.Module.__init__(model)
|
||||
model.model_stage = "cosyvoice3_talker"
|
||||
@@ -124,12 +139,14 @@ def _make_sampling_metadata(
|
||||
|
||||
|
||||
def test_split_request_ids_uses_seq_token_counts():
|
||||
CosyVoice3Model, _ = _cosyvoice3_model_and_runner()
|
||||
ids = torch.tensor([10, 11, 12, 13, 14], dtype=torch.long)
|
||||
chunks = CosyVoice3Model._split_request_ids(ids, [2, 2, 2])
|
||||
assert [c.tolist() for c in chunks] == [[10, 11], [12, 13], [14]]
|
||||
|
||||
|
||||
def test_split_request_ids_honors_single_request_seq_token_counts():
|
||||
CosyVoice3Model, _ = _cosyvoice3_model_and_runner()
|
||||
ids = torch.tensor([10, 11, 12, 13, 14], dtype=torch.long)
|
||||
chunks = CosyVoice3Model._split_request_ids(ids, [3])
|
||||
assert [c.tolist() for c in chunks] == [[10, 11, 12]]
|
||||
@@ -376,6 +393,7 @@ def test_gpu_ar_model_runner_prefers_model_sampler_when_opted_in():
|
||||
def update_async_output_token_ids(self):
|
||||
self.updated = True
|
||||
|
||||
_, GPUARModelRunner = _cosyvoice3_model_and_runner()
|
||||
runner = object.__new__(GPUARModelRunner)
|
||||
runner.input_batch = _DummyInputBatch()
|
||||
runner.model = SimpleNamespace(
|
||||
@@ -407,6 +425,7 @@ def test_gpu_ar_model_runner_supplies_req_output_history_to_model_sampler():
|
||||
def update_async_output_token_ids(self):
|
||||
raise AssertionError("fallback async repair should not run for model sampler path")
|
||||
|
||||
_, GPUARModelRunner = _cosyvoice3_model_and_runner()
|
||||
runner = object.__new__(GPUARModelRunner)
|
||||
runner.input_batch = _DummyInputBatch()
|
||||
runner.model = SimpleNamespace(
|
||||
@@ -446,6 +465,7 @@ def test_gpu_ar_model_runner_repairs_async_placeholders_for_model_sampler():
|
||||
def update_async_output_token_ids(self):
|
||||
raise AssertionError("fallback async repair should not run for model sampler path")
|
||||
|
||||
_, GPUARModelRunner = _cosyvoice3_model_and_runner()
|
||||
runner = object.__new__(GPUARModelRunner)
|
||||
runner.input_batch = _DummyInputBatch()
|
||||
runner.model = SimpleNamespace(
|
||||
|
||||
Reference in new issue
Block a user