Keep release alignment focused and clarify baseline fixes

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>

.buildkite/cuda/rebase-pipeline.yml

@@ -33,13 +33,3 @@ steps:
       - python3 .buildkite/common/scripts/upload_pipeline.py --upload --all .buildkite/cuda/test-nightly.yml
     agents:
       queue: "cpu_queue_premerge"
-
-  # Include the slow model cases moved from nightly into weekly for release validation.
-  - label: "Upload Weekly Pipeline"
-    depends_on: image-build
-    key: upload-weekly-pipeline
-    if: build.env("WEEKLY") == "1"
-    commands:
-      - python3 .buildkite/common/scripts/upload_pipeline.py --upload --all .buildkite/cuda/test-weekly.yml
-    agents:
-      queue: "cpu_queue_premerge"

tests/diffusion/quantization/test_fp8_config.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for the unified quantization framework."""
 
 import pytest

tests/engine/test_arg_utils_shared_fields.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for vllm_omni.engine.arg_utils — invariants that must
 hold for the orchestrator/engine/server CLI flag partition."""
 

tests/engine/test_omni_request.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Regression tests for OmniRequest substitutability with the base Request.
 
 vllm-omni rebinds ``vllm.v1.request.Request`` to ``OmniRequest`` at import

tests/engine/test_omni_request_output.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for OmniRequestOutput class."""
 
 import numpy as np

tests/engine/test_stage_engine_core_client.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for StageEngineCoreClient.check_health()."""
 
 from __future__ import annotations

tests/helpers/distributed.py

@@ -29,7 +29,7 @@ def start_collective_workers(
 def configure_cpu_collective_worker() -> None:
     """Select CPU platforms inside a spawned worker that only uses Gloo.
 
-    Upstream coordinators now resolve local ranks through the selected
+    Upstream coordinators resolve local ranks through the selected
     platform's visible devices. CPU ranks can outnumber visible GPUs, so they
     must use the CPU platform even when the test host has CUDA installed.
     The worker still creates real process groups and runs real collectives.

tests/model_executor/test_cfg_pairing.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Unit tests for the model-neutral CFG request-pairing scheduler patches.
 
 The patches keep a cond/uncond request pair step-locked inside one engine:

vllm_omni/core/sched/utils.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Shared utilities for omni schedulers."""
 
 from vllm.v1.core.kv_cache_manager import KVCacheManager

vllm_omni/model_executor/models/common/cfg_pairing.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 #
 # Adapted from nvidia/Nemotron-Labs-Audex-2B (Apache-2.0):
 #   inference_scripts_vllm/audiogen_scripts/vllm_cfg_patch.py

vllm_omni/model_executor/models/voxcpm2/scheduler.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from __future__ import annotations
 

vllm_omni/model_executor/stage_input_processors/cosyvoice3.py

@@ -305,8 +305,8 @@ _FULL_PAYLOAD_REPLACE_KEYS: frozenset[str] = frozenset(
     }
 )
 
-# vLLM 0.31 keeps the input conditioning snapshot in the request payload and
-# the talker also emits its padded snapshot. Both describe the same prompt;
+# The input conditioning snapshot remains in the request payload and the
+# talker also emits its padded snapshot. Both describe the same prompt;
 # concatenating them creates incompatible shapes before speech_token_len can
 # unpad the emitted value for code2wav.
 for _conditioning_key in _FULL_PAYLOAD_REPLACE_KEYS: