Merge tag 'v0.18.1rc0' into v0.18.1rc0-ori

tests/models/multimodal/processing/test_internvl.py

@@ -23,7 +23,7 @@ def _get_expected_num_patches(
     min_num: int,
     max_num: int,
 ):
-    from vllm.model_executor.models.internvl import (
+    from vllm.transformers_utils.processors.internvl import (
         calculate_internvl_targets,
         get_internvl_target_ratios,
     )

tests/models/multimodal/processing/test_nemotron_vl.py

@@ -23,7 +23,7 @@ def _get_expected_num_patches(
     min_num: int,
     max_num: int,
 ):
-    from vllm.model_executor.models.nemotron_vl import (
+    from vllm.transformers_utils.processors.nemotron_vl import (
         calculate_nemotron_vl_targets,
         get_nemotron_vl_target_ratios,
     )

tests/models/multimodal/processing/test_qwen2_5_omni_embed.py

@@ -185,14 +185,16 @@ def make_mock_model(hidden: int = 8):
 
     # super().embed_input_ids → use SupportsMultiModal.embed_input_ids
     def fake_super_embed(
-        ids, mm_embs=None, *, is_multimodal=None, handle_oov_mm_token=False
+        ids,
+        mm_embs=None,
+        *,
+        is_multimodal=None,
     ):
         return SupportsMultiModal.embed_input_ids(
             model,
             ids,
             mm_embs,
             is_multimodal=is_multimodal,
-            handle_oov_mm_token=handle_oov_mm_token,
         )
 
     # Bind embed_input_ids as the real method

tests/models/quantization/test_mxfp8.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+"""E2E tests for online MXFP8 quantization.
+
+Loads a BF16 model with ``--quantization mxfp8`` (online quantization) and
+compares log-probabilities against the same model served in BF16 without
+quantization.  This exercises the full pipeline: config parsing,
+``Mxfp8OnlineLinearMethod``, ``Mxfp8OnlineMoEMethod``, weight loading,
+online quantization / shuffling, and inference through ``apply_monolithic``.
+
+Layer skipping (``modules_to_not_convert``) is configured in the model's
+``config.json`` under ``quantization_config`` and is not tested here.
+
+``example_prompts`` is a pytest fixture (from conftest.py) that loads 8
+diverse prompts from ``tests/prompts/example.txt``.
+"""
+
+import pytest
+
+from tests.quantization.utils import is_quant_method_supported
+
+from ..utils import check_logprobs_close
+
+# A small MoE model that fits on a single GPU and has both linear + MoE layers.
+MOE_MODEL = "Qwen/Qwen3-30B-A3B"
+# A small dense model (no MoE) to validate the linear-only path.
+DENSE_MODEL = "Qwen/Qwen3-0.6B"
+
+MAX_MODEL_LEN = 1024
+MAX_TOKENS = 4
+NUM_LOG_PROBS = 8
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("mxfp8"),
+    reason="mxfp8 is not supported on this GPU type (requires sm_100+).",
+)
+@pytest.mark.quant_model
+@pytest.mark.parametrize("model", [DENSE_MODEL, MOE_MODEL], ids=["dense", "moe"])
+def test_mxfp8_logprobs(
+    vllm_runner,
+    example_prompts,
+    model: str,
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    """Compare BF16 baseline logprobs against online MXFP8-quantized model.
+
+    Runs the same model twice -- once in BF16 (baseline) and once with
+    online MXFP8 quantization -- then checks that the top log-probabilities
+    are close.  Only 4 tokens are generated to keep the test fast while
+    still catching numerical divergence.
+    """
+    with monkeypatch.context() as m:
+        m.setenv("TOKENIZERS_PARALLELISM", "true")
+
+        with vllm_runner(
+            model,
+            max_model_len=MAX_MODEL_LEN,
+            enforce_eager=True,
+        ) as vllm_model:
+            baseline_outputs = vllm_model.generate_greedy_logprobs(
+                example_prompts, MAX_TOKENS, NUM_LOG_PROBS
+            )
+
+        with vllm_runner(
+            model,
+            max_model_len=MAX_MODEL_LEN,
+            enforce_eager=True,
+            quantization="mxfp8",
+        ) as vllm_model:
+            test_outputs = vllm_model.generate_greedy_logprobs(
+                example_prompts, MAX_TOKENS, NUM_LOG_PROBS
+            )
+
+        check_logprobs_close(
+            outputs_0_lst=baseline_outputs,
+            outputs_1_lst=test_outputs,
+            name_0="bf16",
+            name_1="mxfp8",
+        )
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("mxfp8"),
+    reason="mxfp8 is not supported on this GPU type (requires sm_100+).",
+)
+@pytest.mark.quant_model
+@pytest.mark.parametrize("model", [DENSE_MODEL, MOE_MODEL], ids=["dense", "moe"])
+def test_mxfp8_generation(vllm_runner, model: str) -> None:
+    """Smoke test: verify online MXFP8 model generates coherent text."""
+    prompt = "1 2 3 4 5"
+    with vllm_runner(
+        model,
+        enforce_eager=True,
+        quantization="mxfp8",
+        max_model_len=MAX_MODEL_LEN,
+    ) as vllm_model:
+        output = vllm_model.generate_greedy([prompt], max_tokens=5)
+
+    generated = output[0][1]
+    assert len(generated) > len(prompt), (
+        f"MXFP8 model produced no new tokens. Output: {generated!r}"
+    )

tests/models/registry.py

@@ -628,6 +628,11 @@ _LATE_INTERACTION_EXAMPLE_MODELS = {
         trust_remote_code=True,
         hf_overrides={"architectures": ["ColBERTJinaRobertaModel"]},
     ),
+    "ColBERTLfm2Model": _HfExamplesInfo(
+        "LiquidAI/LFM2-ColBERT-350M",
+        trust_remote_code=True,
+        hf_overrides={"architectures": ["ColBERTLfm2Model"]},
+    ),
     # [Multimodal]
     "ColModernVBertForRetrieval": _HfExamplesInfo(
         "ModernVBERT/colmodernvbert-merged",
@@ -639,6 +644,11 @@ _LATE_INTERACTION_EXAMPLE_MODELS = {
     "OpsColQwen3Model": _HfExamplesInfo(
         "OpenSearch-AI/Ops-Colqwen3-4B", trust_remote_code=True
     ),
+    "ColQwen3_5": _HfExamplesInfo(
+        "athrael-soju/colqwen3.5-4.5B-v3",
+        trust_remote_code=True,
+        max_model_len=4096,
+    ),
     "Qwen3VLNemotronEmbedModel": _HfExamplesInfo(
         "nvidia/nemotron-colembed-vl-4b-v2",
     ),
@@ -774,7 +784,8 @@ _MULTIMODAL_EXAMPLE_MODELS = {
         "rednote-hilab/dots.ocr", trust_remote_code=True
     ),
     "Eagle2_5_VLForConditionalGeneration": _HfExamplesInfo(
-        "nvidia/Eagle2.5-8B", trust_remote_code=True, is_available_online=False
+        "nvidia/Eagle2.5-8B",
+        trust_remote_code=True,
     ),
     "Emu3ForConditionalGeneration": _HfExamplesInfo("BAAI/Emu3-Chat-hf"),
     "Ernie4_5_VLMoeForConditionalGeneration": _HfExamplesInfo(
@@ -1116,6 +1127,11 @@ _MULTIMODAL_EXAMPLE_MODELS = {
         tokenizer_mode="mistral",
     ),
     # [Encoder-decoder]
+    "CohereASRForConditionalGeneration": _HfExamplesInfo(
+        "/host/engines/vllm/audio/2b-release",
+        trust_remote_code=True,
+        is_available_online=False,  # TODO (ekagra): revert after asr release
+    ),
     "NemotronParseForConditionalGeneration": _HfExamplesInfo(
         "nvidia/NVIDIA-Nemotron-Parse-v1.1", trust_remote_code=True
     ),

tests/models/test_terratorch.py

@@ -8,7 +8,7 @@ from tests.conftest import VllmRunner
 from tests.utils import create_new_process_for_each_test
 
 
-@create_new_process_for_each_test()  # Memory is not cleaned up properly otherwise
+@create_new_process_for_each_test()  # Hangs otherwise
 @pytest.mark.parametrize(
     "model",
     [

tests/multimodal/media/test_audio.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import base64
 from pathlib import Path
 from unittest.mock import patch
 
 import librosa
 import numpy as np
+import pybase64 as base64
 import pytest
 
 from vllm.multimodal.media import AudioMediaIO
 
+from ...conftest import AudioTestAssets
+
 pytestmark = pytest.mark.cpu_test
 
 ASSETS_DIR = Path(__file__).parent.parent / "assets"
@@ -22,40 +24,32 @@ def dummy_audio():
 
 
 @pytest.fixture
-def dummy_audio_bytes():
-    return b"FAKEAUDIOBYTES"
+def dummy_audio_bytes(audio_assets: AudioTestAssets):
+    with open(audio_assets[0].get_local_path(), "rb") as f:
+        return f.read()
 
 
 def test_audio_media_io_load_bytes(dummy_audio_bytes):
     audio_io = AudioMediaIO()
-    with patch("librosa.load") as mock_load:
-        mock_load.return_value = (np.array([0.1, 0.2]), 16000)
-        out = audio_io.load_bytes(dummy_audio_bytes)
-        mock_load.assert_called_once()
-        assert isinstance(out[0], np.ndarray)
-        assert out[1] == 16000
+    out = audio_io.load_bytes(dummy_audio_bytes)
+    assert isinstance(out[0], np.ndarray)
+    assert out[1] == 16000
 
 
 def test_audio_media_io_load_base64(dummy_audio_bytes):
     audio_io = AudioMediaIO()
     encoded = base64.b64encode(dummy_audio_bytes).decode("utf-8")
-    with patch.object(AudioMediaIO, "load_bytes") as mock_load_bytes:
-        mock_load_bytes.return_value = (np.array([0.1, 0.2]), 16000)
-        out = audio_io.load_base64("audio/wav", encoded)
-        mock_load_bytes.assert_called_once()
-        assert isinstance(out[0], np.ndarray)
-        assert out[1] == 16000
+    out = audio_io.load_base64("audio/wav", encoded)
+    assert isinstance(out[0], np.ndarray)
+    assert out[1] == 16000
 
 
-def test_audio_media_io_load_file():
+def test_audio_media_io_load_file(audio_assets: AudioTestAssets):
     audio_io = AudioMediaIO()
-    path = Path("/fake/path.wav")
-    with patch("librosa.load") as mock_load:
-        mock_load.return_value = (np.array([0.1, 0.2]), 16000)
-        out = audio_io.load_file(path)
-        mock_load.assert_called_once_with(path, sr=None)
-        assert isinstance(out[0], np.ndarray)
-        assert out[1] == 16000
+    path = audio_assets[0].get_local_path()
+    out = audio_io.load_file(path)
+    assert isinstance(out[0], np.ndarray)
+    assert out[1] == 16000
 
 
 def test_audio_media_io_encode_base64(dummy_audio):

tests/multimodal/media/test_connector.py

@@ -2,13 +2,13 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import asyncio
-import base64
 import mimetypes
 import os
 from tempfile import NamedTemporaryFile, TemporaryDirectory
 
 import aiohttp
 import numpy as np
+import pybase64 as base64
 import pytest
 import requests
 import torch

tests/multimodal/media/test_video.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import io
 from pathlib import Path
 
 import numpy as np
 import numpy.typing as npt
+import pybase64
 import pytest
 from PIL import Image
 
@@ -235,3 +237,53 @@ def test_video_media_io_backend_env_var_fallback(monkeypatch: pytest.MonkeyPatch
         frames_missing, metadata_missing = videoio_missing.load_bytes(b"test")
         np.testing.assert_array_equal(frames_missing, FAKE_OUTPUT_2)
         assert metadata_missing["video_backend"] == "test_video_backend_override_2"
+
+
+def test_load_base64_jpeg_returns_metadata():
+    """Regression test: load_base64 with video/jpeg must return metadata.
+
+    Previously, base64 JPEG frame sequences returned an empty dict for
+    metadata, which broke downstream consumers that rely on fields like
+    total_num_frames and fps. See PR #37301.
+    """
+
+    num_test_frames = 3
+    frame_width, frame_height = 8, 8
+
+    # Build a few tiny JPEG frames and base64-encode them
+    b64_frames = []
+    for i in range(num_test_frames):
+        img = Image.new("RGB", (frame_width, frame_height), color=(i * 80, 0, 0))
+        buf = io.BytesIO()
+        img.save(buf, format="JPEG")
+        b64_frames.append(pybase64.b64encode(buf.getvalue()).decode("ascii"))
+
+    data = ",".join(b64_frames)
+
+    imageio = ImageMediaIO()
+    videoio = VideoMediaIO(imageio, num_frames=num_test_frames)
+    frames, metadata = videoio.load_base64("video/jpeg", data)
+
+    # Frames array shape: (num_frames, H, W, 3)
+    assert frames.shape[0] == num_test_frames
+
+    # All required metadata keys must be present
+    required_keys = {
+        "total_num_frames",
+        "fps",
+        "duration",
+        "video_backend",
+        "frames_indices",
+        "do_sample_frames",
+    }
+    assert required_keys.issubset(metadata.keys()), (
+        f"Missing metadata keys: {required_keys - metadata.keys()}"
+    )
+
+    assert metadata["total_num_frames"] == num_test_frames
+    assert metadata["video_backend"] == "jpeg_sequence"
+    assert metadata["frames_indices"] == list(range(num_test_frames))
+    assert metadata["do_sample_frames"] is False
+    # Default fps=1 → duration == num_frames
+    assert metadata["fps"] == 1.0
+    assert metadata["duration"] == float(num_test_frames)

tests/multimodal/test_audio.py

@@ -14,7 +14,7 @@ from vllm.multimodal.audio import (
     AudioSpec,
     ChannelReduction,
     normalize_audio,
-    resample_audio_librosa,
+    resample_audio_pyav,
     resample_audio_scipy,
     split_audio,
 )
@@ -25,14 +25,14 @@ def dummy_audio():
     return np.array([0.0, 0.1, 0.2, 0.3, 0.4], dtype=float)
 
 
-def test_resample_audio_librosa(dummy_audio):
-    with patch("vllm.multimodal.audio.librosa.resample") as mock_resample:
-        mock_resample.return_value = dummy_audio * 2
-        out = resample_audio_librosa(dummy_audio, orig_sr=44100, target_sr=22050)
-        mock_resample.assert_called_once_with(
-            dummy_audio, orig_sr=44100, target_sr=22050
-        )
-        assert np.all(out == dummy_audio * 2)
+def test_resample_audio_pyav(dummy_audio):
+    out_down = resample_audio_pyav(dummy_audio, orig_sr=4, target_sr=2)
+    out_up = resample_audio_pyav(dummy_audio, orig_sr=2, target_sr=4)
+    out_same = resample_audio_pyav(dummy_audio, orig_sr=4, target_sr=4)
+
+    assert len(out_down) == 3
+    assert len(out_up) == 10
+    assert np.all(out_same == dummy_audio)
 
 
 def test_resample_audio_scipy(dummy_audio):
@@ -56,9 +56,9 @@ def test_resample_audio_scipy_non_integer_ratio(dummy_audio):
     assert np.isfinite(out).all()
 
 
-def test_audio_resampler_librosa_calls_resample(dummy_audio):
-    resampler = AudioResampler(target_sr=22050, method="librosa")
-    with patch("vllm.multimodal.audio.resample_audio_librosa") as mock_resample:
+def test_audio_resampler_pyav_calls_resample(dummy_audio):
+    resampler = AudioResampler(target_sr=22050, method="pyav")
+    with patch("vllm.multimodal.audio.resample_audio_pyav") as mock_resample:
         mock_resample.return_value = dummy_audio
         out = resampler.resample(dummy_audio, orig_sr=44100)
         mock_resample.assert_called_once_with(
@@ -423,13 +423,13 @@ class TestAudioPipelineE2E:
         # Verify channel averaging: mean of [0.5, -0.5] = 0.0
         np.testing.assert_array_almost_equal(audio_output, np.zeros(16000), decimal=5)
 
-    def test_librosa_mono_passthrough_e2e(self):
-        """Full pipeline: librosa mono format → preserved as mono."""
+    def test_pyav_mono_passthrough_e2e(self):
+        """Full pipeline: pyav mono format → preserved as mono."""
         from vllm.multimodal.parse import MultiModalDataParser
 
-        # Simulate librosa output: already mono (time,) format
-        mono_librosa = np.random.randn(16000).astype(np.float32)
-        assert mono_librosa.shape == (16000,)
+        # Simulate pyav output: already mono (time,) format
+        mono_pyav = np.random.randn(16000).astype(np.float32)
+        assert mono_pyav.shape == (16000,)
 
         # Create parser with mono normalization
         parser = MultiModalDataParser(
@@ -438,7 +438,7 @@ class TestAudioPipelineE2E:
         )
 
         # Process audio through the parser
-        result = parser._parse_audio_data((mono_librosa, 16000))
+        result = parser._parse_audio_data((mono_pyav, 16000))
         audio_output = result.get(0)
 
         # Verify output is still mono 1D
@@ -446,7 +446,7 @@ class TestAudioPipelineE2E:
         assert audio_output.shape == (16000,)
 
         # Verify audio content is preserved
-        np.testing.assert_array_almost_equal(audio_output, mono_librosa)
+        np.testing.assert_array_almost_equal(audio_output, mono_pyav)
 
     def test_multichannel_5_1_surround_to_mono_e2e(self):
         """Full pipeline: 5.1 surround (6 channels) → mono output."""

tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/sparse_embeddings_processor.py

@@ -3,10 +3,10 @@
 
 from collections.abc import Sequence
 
-from vllm.config import VllmConfig
+from vllm.config import ModelConfig, PoolerConfig, VllmConfig
 from vllm.entrypoints.openai.engine.protocol import UsageInfo
+from vllm.entrypoints.pooling.base.protocol import EmbedRequestMixin
 from vllm.inputs.data import PromptType
-from vllm.logger import init_logger
 from vllm.outputs import PoolingRequestOutput
 from vllm.plugins.io_processors.interface import (
     IOProcessor,
@@ -16,14 +16,13 @@ from vllm.renderers import BaseRenderer
 from vllm.tokenizers.detokenizer_utils import convert_ids_list_to_tokens
 
 from .types import (
+    EMBED_TASKS,
     SparseEmbeddingCompletionRequestMixin,
     SparseEmbeddingResponse,
     SparseEmbeddingResponseData,
     SparseEmbeddingTokenWeight,
 )
 
-logger = init_logger(__name__)
-
 
 class BgeM3SparseEmbeddingsProcessor(
     IOProcessor[SparseEmbeddingCompletionRequestMixin, SparseEmbeddingResponse]
@@ -33,6 +32,22 @@ class BgeM3SparseEmbeddingsProcessor(
         self.offline_requests: list[SparseEmbeddingCompletionRequestMixin] = []
         self.online_requests: dict[str, SparseEmbeddingCompletionRequestMixin] = {}
         self.renderer: BaseRenderer = renderer
+        self.default_pooling_params = {}
+        pooler_config: PoolerConfig = vllm_config.model_config.pooler_config
+        if pooler_config is not None:
+            for param in ["use_activation", "dimensions"]:
+                if getattr(pooler_config, param, None) is None:
+                    continue
+                self.default_pooling_params[param] = getattr(pooler_config, param)
+        self.embed_dimensions = vllm_config.model_config.embedding_size
+        self.embed_request_queue: list[EmbedRequestMixin] = []
+
+    def __repr__(self) -> str:
+        return (
+            f"BgeM3SparseEmbeddingsProcessor("
+            f"embed_dimensions={self.embed_dimensions}, "
+            f"default_pooling_params={self.default_pooling_params})"
+        )
 
     def merge_pooling_params(
         self,
@@ -41,7 +56,57 @@ class BgeM3SparseEmbeddingsProcessor(
         if params is None:
             params = PoolingParams()
         # refer to PoolingCompletionRequest.to_pooling_params
-        params.task = "token_classify"
+        # set and verify pooling params
+        params.skip_reading_prefix_cache = True
+
+        raw_embed_request = self.embed_request_queue.pop(0)
+        if raw_embed_request.embed_task not in EMBED_TASKS:
+            raise ValueError(
+                f"Unsupported task {raw_embed_request}, "
+                f"Supported tasks are {EMBED_TASKS}"
+            )
+        has_dense_embed = True
+        if raw_embed_request.embed_task == "dense":
+            params.task = "embed"
+            params.skip_reading_prefix_cache = False
+        elif raw_embed_request.embed_task == "sparse":
+            params.task = "token_classify"
+            has_dense_embed = False
+        else:
+            params.task = "embed&token_classify"
+        params.use_activation = raw_embed_request.use_activation
+        if params.use_activation is None:
+            params.use_activation = True
+        if not has_dense_embed:
+            params.dimensions = None
+            return params
+
+        params.dimensions = raw_embed_request.dimensions
+
+        model_config: ModelConfig = self.vllm_config.model_config
+        for param in self.default_pooling_params:
+            if getattr(params, param, None) is None:
+                setattr(params, param, self.default_pooling_params[param])
+
+        if params.dimensions is not None:
+            if not model_config.is_matryoshka:
+                raise ValueError(
+                    f'Model "{model_config.served_model_name}" does not '
+                    f"support matryoshka representation, "
+                    f"changing output dimensions will lead to poor results."
+                )
+
+            mds = model_config.matryoshka_dimensions
+            if mds is not None:
+                if params.dimensions not in mds:
+                    raise ValueError(
+                        f"Model {model_config.served_model_name!r} "
+                        f"only supports {str(mds)} matryoshka dimensions, "
+                        f"use other output dimensions will "
+                        f"lead to poor results."
+                    )
+            elif params.dimensions < 1:
+                raise ValueError("Dimensions must be greater than 0")
         return params
 
     def parse_request(
@@ -61,14 +126,16 @@ class BgeM3SparseEmbeddingsProcessor(
         if request_id is not None:
             assert request_id not in self.online_requests, "request_id duplicated"
             self.online_requests[request_id] = prompt
+            self.embed_request_queue.extend(prompt.to_embed_requests_online())
         else:
             self.offline_requests.append(prompt)
+            self.embed_request_queue.extend(prompt.to_embed_requests_offline())
         return prompt.input
 
     def _get_sparse_embedding_request(self, request_id: str | None = None):
         if request_id:
             return self.online_requests.pop(request_id, None)
-        return self.offline_requests.pop()
+        return self.offline_requests.pop(0)
 
     def _build_sparse_embedding_token_weights(
         self,
@@ -100,26 +167,45 @@ class BgeM3SparseEmbeddingsProcessor(
     ) -> SparseEmbeddingResponse:
         num_prompt_tokens = 0
         response_data = []
-        return_tokens = self._get_sparse_embedding_request(request_id).return_tokens
+        raw_request = self._get_sparse_embedding_request(request_id)
+        has_dense_embed = raw_request.embed_task in ["dense", "dense&sparse"]
+        has_sparse_embed = raw_request.embed_task in ["sparse", "dense&sparse"]
+        embed_dimensions = 0
+        if has_dense_embed:
+            embed_dimensions = (
+                self.embed_dimensions
+                if raw_request.dimensions is None
+                else raw_request.dimensions
+            )
         for idx in range(len(model_output)):
             mo = model_output[idx]
-            sparse_embedding: dict[int, float] = {}
+            sparse_embedding_dict: dict[int, float] = {}
             num_prompt_tokens += len(mo.prompt_token_ids)
-            if len(mo.prompt_token_ids) != len(mo.outputs.data):
-                # this is the case that add_special_tokens is True,
-                # which means first token and last token are special tokens
-                mo.prompt_token_ids = mo.prompt_token_ids[1:]
-            for token_id, weight in zip(mo.prompt_token_ids, mo.outputs.data.tolist()):
-                sparse_embedding[token_id] = max(
-                    weight, sparse_embedding.get(token_id, 0.0)
+            dense_embedding: list[float] | None = None
+            sparse_embedding: list[SparseEmbeddingTokenWeight] | None = None
+            if has_dense_embed:
+                dense_embedding = mo.outputs.data[:embed_dimensions].tolist()
+            if has_sparse_embed:
+                sparse_weights = mo.outputs.data[embed_dimensions:].tolist()
+                if len(mo.prompt_token_ids) != len(sparse_weights):
+                    # this is the case that add_special_tokens is True,
+                    # which means first token and last token are special tokens
+                    mo.prompt_token_ids = mo.prompt_token_ids[1:]
+                for token_id, weight in zip(mo.prompt_token_ids, sparse_weights):
+                    sparse_embedding_dict[token_id] = max(
+                        weight, sparse_embedding_dict.get(token_id, 0.0)
+                    )
+                sparse_embedding = self._build_sparse_embedding_token_weights(
+                    sparse_embedding_dict,
+                    raw_request.return_tokens,
                 )
+
             response_data.append(
                 SparseEmbeddingResponseData(
                     index=idx,
-                    sparse_embedding=self._build_sparse_embedding_token_weights(
-                        sparse_embedding,
-                        return_tokens,
-                    ),
+                    object=raw_request.embed_task,
+                    sparse_embedding=sparse_embedding,
+                    dense_embedding=dense_embedding,
                 )
             )
 

tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/types.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
+from typing import Literal, get_args
+
 from pydantic import BaseModel, Field
 
 from vllm.entrypoints.openai.engine.protocol import UsageInfo
-from vllm.entrypoints.pooling.base.protocol import CompletionRequestMixin
+from vllm.entrypoints.pooling.base.protocol import (
+    CompletionRequestMixin,
+    EmbedRequestMixin,
+)
+
+EmbedTask = Literal[
+    "sparse",
+    "dense",
+    "dense&sparse",
+]
+
+EMBED_TASKS: tuple[EmbedTask, ...] = get_args(EmbedTask)
 
 
-class SparseEmbeddingCompletionRequestMixin(CompletionRequestMixin):
+class SparseEmbeddingCompletionRequestMixin(CompletionRequestMixin, EmbedRequestMixin):
     return_tokens: bool | None = Field(
         default=None,
         description="Whether to return dict shows the mapping of token_id to text."
         "`None` or False means not return.",
     )
+    embed_task: EmbedTask = Field(
+        default="dense&sparse",
+        description="embed task, can be one of 'sparse', 'dense' , 'dense&sparse', "
+        "default to 'dense&sparse'",
+    )
+
+    def to_embed_requests_offline(self) -> list[EmbedRequestMixin]:
+        if isinstance(self.input, list):
+            return [self] * len(self.input)
+        return [self]
+
+    def to_embed_requests_online(self) -> list[EmbedRequestMixin]:
+        return [self]
 
 
 class SparseEmbeddingTokenWeight(BaseModel):
@@ -23,8 +49,9 @@ class SparseEmbeddingTokenWeight(BaseModel):
 
 class SparseEmbeddingResponseData(BaseModel):
     index: int
-    object: str = "sparse-embedding"
-    sparse_embedding: list[SparseEmbeddingTokenWeight]
+    object: str = "dense&sparse"
+    sparse_embedding: list[SparseEmbeddingTokenWeight] | None
+    dense_embedding: list[float] | None
 
 
 class SparseEmbeddingResponse(BaseModel):

tests/plugins/prithvi_io_processor_plugin/prithvi_io_processor/prithvi_processor.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-import base64
 import datetime
 import os
 import tempfile
@@ -11,6 +10,7 @@ from typing import Any
 
 import albumentations
 import numpy as np
+import pybase64 as base64
 import rasterio
 import regex as re
 import torch

tests/plugins_tests/test_bge_m3_sparse_io_processor_plugins.py

@@ -19,6 +19,12 @@ model_config = {
     ),
 }
 
+dense_embedding_sum = [
+    -0.7214539647102356,  # "What is the capital of France?"
+    -0.6926871538162231,  # "What is the capital of Germany?"
+    -0.7129564881324768,  # "What is the capital of Spain?"
+]
+
 
 def _float_close(expected: object, result: object):
     assert isinstance(expected, float) and isinstance(result, float), (
@@ -33,6 +39,12 @@ def _get_attr_or_val(obj: object | dict, key: str):
     return getattr(obj, key, None)
 
 
+def _check_dense_embedding(data, index=0):
+    assert _float_close(sum(data), dense_embedding_sum[index]), (
+        "dense-embedding result not match"
+    )
+
+
 def _check_sparse_embedding(data, check_tokens=False):
     expected_weights = [
         {"token_id": 32, "weight": 0.0552978515625, "token": "?"},
@@ -109,7 +121,7 @@ async def test_bge_m3_sparse_plugin_online(
     assert len(_get_attr_or_val(parsed_response, "data")) > 0
 
     data_entry = _get_attr_or_val(parsed_response, "data")[0]
-    assert _get_attr_or_val(data_entry, "object") == "sparse-embedding"
+    assert _get_attr_or_val(data_entry, "object") == "dense&sparse"
     assert _get_attr_or_val(data_entry, "sparse_embedding")
 
     # Verify sparse embedding format
@@ -117,6 +129,11 @@ async def test_bge_m3_sparse_plugin_online(
     assert isinstance(sparse_embedding, list)
     _check_sparse_embedding(sparse_embedding, return_tokens)
 
+    # Verify dense embedding format
+    dense_embedding = _get_attr_or_val(data_entry, "dense_embedding")
+    assert isinstance(dense_embedding, list)
+    _check_dense_embedding(dense_embedding)
+
     # Verify usage information
     usage = _get_attr_or_val(parsed_response, "usage")
     assert usage, f"usage not found for {parsed_response}"
@@ -164,6 +181,9 @@ def test_bge_m3_sparse_plugin_offline(vllm_runner, return_tokens: bool):
         sparse_embedding = output.sparse_embedding
         assert isinstance(sparse_embedding, list)
         _check_sparse_embedding(sparse_embedding, return_tokens)
+        dense_embedding = output.dense_embedding
+        assert isinstance(dense_embedding, list)
+        _check_dense_embedding(dense_embedding)
 
     # Verify usage
     assert response.usage.prompt_tokens > 0
@@ -206,6 +226,9 @@ def test_bge_m3_sparse_plugin_offline_multiple_inputs(vllm_runner):
         # Each output should have sparse embeddings
         sparse_embedding = output.sparse_embedding
         assert isinstance(sparse_embedding, list)
+        dense_embedding = output.dense_embedding
+        assert isinstance(dense_embedding, list)
+        _check_dense_embedding(dense_embedding, i)
 
     # Verify usage
     assert response.usage.prompt_tokens > 0

tests/plugins_tests/test_terratorch_io_processor_plugins.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import base64
 import io
 
 import imagehash
+import pybase64 as base64
 import pytest
 import requests
 from PIL import Image

tests/quantization/test_mi3xx_moe.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+
+def test_mi3xx_moe():
+    print("TODO: add tests for Mi3xx MoE quantization")

tests/reasoning/test_kimi_k2_reasoning_parser.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import pytest
+
+from vllm.entrypoints.openai.chat_completion.protocol import ChatCompletionRequest
+from vllm.entrypoints.openai.engine.protocol import DeltaMessage
+from vllm.reasoning.identity_reasoning_parser import IdentityReasoningParser
+from vllm.reasoning.kimi_k2_reasoning_parser import KimiK2ReasoningParser
+from vllm.tokenizers import get_tokenizer
+
+REASONING_MODEL_NAME = "moonshotai/Kimi-K2.5"
+
+
+@pytest.fixture(scope="module")
+def kimi_k2_tokenizer():
+    return get_tokenizer(tokenizer_name=REASONING_MODEL_NAME, trust_remote_code=True)
+
+
+def test_parser_selection_thinking_enabled(kimi_k2_tokenizer):
+    parser = KimiK2ReasoningParser(
+        kimi_k2_tokenizer, chat_template_kwargs={"thinking": True}
+    )
+    assert parser._identity_parser is None
+
+
+def test_parser_selection_thinking_disabled(kimi_k2_tokenizer):
+    parser = KimiK2ReasoningParser(
+        kimi_k2_tokenizer, chat_template_kwargs={"thinking": False}
+    )
+    assert isinstance(parser._identity_parser, IdentityReasoningParser)
+
+
+def test_extract_reasoning_with_think_tags(kimi_k2_tokenizer):
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+    request = ChatCompletionRequest(model="test-model", messages=[], temperature=1.0)
+
+    reasoning, content = parser.extract_reasoning(
+        "<think>step by step reasoning</think>final answer", request
+    )
+    assert reasoning == "step by step reasoning"
+    assert content == "final answer"
+
+
+def test_extract_reasoning_empty_thinking(kimi_k2_tokenizer):
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+    request = ChatCompletionRequest(model="test-model", messages=[], temperature=1.0)
+
+    reasoning, content = parser.extract_reasoning(
+        "<think></think>final answer", request
+    )
+    assert reasoning == ""
+    assert content == "final answer"
+
+
+def test_extract_reasoning_implicit_start(kimi_k2_tokenizer):
+    """When there's no <think> tag, everything is treated as reasoning."""
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+    request = ChatCompletionRequest(model="test-model", messages=[], temperature=1.0)
+
+    reasoning, content = parser.extract_reasoning(
+        "implicit reasoning with no tags", request
+    )
+    assert reasoning == "implicit reasoning with no tags"
+    assert content is None
+
+
+def test_extract_reasoning_tool_section_ends_reasoning(kimi_k2_tokenizer):
+    """<|tool_calls_section_begin|> implicitly ends reasoning."""
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+    request = ChatCompletionRequest(model="test-model", messages=[], temperature=1.0)
+
+    text = "some reasoning<|tool_calls_section_begin|>tool call data"
+    reasoning, content = parser.extract_reasoning(text, request)
+    assert reasoning == "some reasoning"
+    assert content == "<|tool_calls_section_begin|>tool call data"
+
+
+def test_streaming_reasoning_then_content(kimi_k2_tokenizer):
+    """Token-by-token streaming: reasoning tokens then content after </think>."""
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+
+    think_id = parser._start_token_id
+    end_think_id = parser._end_token_id
+    # Use a real token ID from the tokenizer for regular content
+    regular_id = kimi_k2_tokenizer.encode("hello", add_special_tokens=False)[0]
+
+    # First token: <think> — single special token should be skipped
+    result = parser.extract_reasoning_streaming(
+        previous_text="",
+        current_text="<think>",
+        delta_text="<think>",
+        previous_token_ids=[],
+        current_token_ids=[think_id],
+        delta_token_ids=[think_id],
+    )
+    assert result is None
+
+    # Reasoning token
+    result = parser.extract_reasoning_streaming(
+        previous_text="<think>",
+        current_text="<think>step one",
+        delta_text="step one",
+        previous_token_ids=[think_id],
+        current_token_ids=[think_id, regular_id],
+        delta_token_ids=[regular_id],
+    )
+    assert isinstance(result, DeltaMessage)
+    assert result.reasoning == "step one"
+    assert result.content is None
+
+    # End token </think> as single token — should be skipped
+    result = parser.extract_reasoning_streaming(
+        previous_text="<think>step one",
+        current_text="<think>step one</think>",
+        delta_text="</think>",
+        previous_token_ids=[think_id, regular_id],
+        current_token_ids=[think_id, regular_id, end_think_id],
+        delta_token_ids=[end_think_id],
+    )
+    assert result is None
+
+    # Content after </think>
+    content_id = kimi_k2_tokenizer.encode("world", add_special_tokens=False)[0]
+    result = parser.extract_reasoning_streaming(
+        previous_text="<think>step one</think>",
+        current_text="<think>step one</think>answer",
+        delta_text="answer",
+        previous_token_ids=[think_id, regular_id, end_think_id],
+        current_token_ids=[think_id, regular_id, end_think_id, content_id],
+        delta_token_ids=[content_id],
+    )
+    assert isinstance(result, DeltaMessage)
+    assert result.content == "answer"
+
+
+def test_streaming_tool_section_ends_reasoning(kimi_k2_tokenizer):
+    """<|tool_calls_section_begin|> in delta ends reasoning during streaming."""
+    parser = KimiK2ReasoningParser(kimi_k2_tokenizer)
+
+    think_id = parser._start_token_id
+    tool_begin_id = parser._tool_section_start_token_id
+    regular_id = kimi_k2_tokenizer.encode("hello", add_special_tokens=False)[0]
+
+    # Tool section token arrives — should transition from reasoning to content
+    result = parser.extract_reasoning_streaming(
+        previous_text="<think>thinking",
+        current_text="<think>thinking<|tool_calls_section_begin|>",
+        delta_text="<|tool_calls_section_begin|>",
+        previous_token_ids=[think_id, regular_id],
+        current_token_ids=[think_id, regular_id, tool_begin_id],
+        delta_token_ids=[tool_begin_id],
+    )
+    assert isinstance(result, DeltaMessage)
+    assert result.content == "<|tool_calls_section_begin|>"

tests/reasoning/test_step3p5_reasoning_parser.py

@@ -21,119 +21,119 @@ def step3p5_tokenizer():
 
 SIMPLE_REASONING = {
     "output": "This is a reasoning section</think>This is the rest",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 # need to get into parser again to remove newline after </think>
 COMPLETE_REASONING = {
     "output": "This is a reasoning section</think>",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": None,
     "is_reasoning_end": False,
 }
 NO_CONTENT = {
     "output": "This is content",
-    "reasoning_content": "This is content",
+    "reasoning": "This is content",
     "content": None,
     "is_reasoning_end": False,
 }
 NO_REASONING_STREAMING = {
     "output": "This is a reasoning section",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": None,
     "is_reasoning_end": False,
 }
 MULTIPLE_LINES = {
     "output": "This\nThat</think>This is the rest\nThat",
-    "reasoning_content": "This\nThat",
+    "reasoning": "This\nThat",
     "content": "This is the rest\nThat",
     "is_reasoning_end": True,
 }
 SHORTEST_REASONING_NO_STREAMING = {
     "output": "</think>This is the rest",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 SHORTEST_REASONING = {
     "output": "</think>This is the rest",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 REASONING_WITH_THINK = {
     "output": "<think>This is a reasoning section</think>This is the rest",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 COMPLETE_REASONING_WITH_THINK = {
     "output": "<think>This is a reasoning section</think>",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": None,
     "is_reasoning_end": False,
 }
 MULTIPLE_LINES_WITH_THINK = {
     "output": "<think>This\nThat</think>This is the rest\nThat",
-    "reasoning_content": "This\nThat",
+    "reasoning": "This\nThat",
     "content": "This is the rest\nThat",
     "is_reasoning_end": True,
 }
 SHORTEST_REASONING_NO_STREAMING_WITH_THINK = {
     "output": "</think>This is the rest",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 SHORTEST_REASONING_WITH_THINK = {
     "output": "</think>This is the rest",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 THINK_NO_END = {
     "output": "<think>This is a reasoning section",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": None,
     "is_reasoning_end": False,
 }
 EMPTY = {
     "output": "",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": None,
     "is_reasoning_end": False,
 }
 EMPTY_STREAMING = {
     "output": "",
-    "reasoning_content": None,
+    "reasoning": None,
     "content": None,
     "is_reasoning_end": False,
 }
 NEW_LINE = {
     "output": "\n<think>This is a reasoning section</think>\nThis is the rest",
-    "reasoning_content": "This is a reasoning section",
+    "reasoning": "This is a reasoning section",
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 
 NEW_LINE_STREAMING = {
     "output": "\n<think>This is a reasoning section\n</think>\nThis is the rest",
-    "reasoning_content": "\nThis is a reasoning section",
+    "reasoning": "\nThis is a reasoning section",
     "content": "This is the rest",
     "is_reasoning_end": True,
 }
 
 NEW_LINE_STREAMING_COMPLEX_CONTENT = {
     "output": "\n This is a \n reasoning section\n\n\n</think>\n\nThis is the rest",
-    "reasoning_content": "\n This is a \n reasoning section\n\n",
+    "reasoning": "\n This is a \n reasoning section\n\n",
     "content": "\nThis is the rest",
     "is_reasoning_end": True,
 }
 
 MULTI_TURN_PROMPT_CONTENT = {
     "output": "<think> This is last turn's reasoning section </think> hello <think>",
-    "reasoning_content": "",
+    "reasoning": "",
     "content": "",
     "is_reasoning_end": False,
 }
@@ -296,7 +296,7 @@ def test_reasoning(
     print(f"content: {content}")
     test_id = request.node.callspec.id if hasattr(request.node, "callspec") else None
     if request.node.callspec.id != "multi_turn_prompt_content":
-        assert reasoning == param_dict["reasoning_content"]
+        assert reasoning == param_dict["reasoning"]
         assert content == param_dict["content"]
 
     # Test is_reasoning_end

tests/renderers/test_sparse_tensor_validation.py

@@ -5,9 +5,9 @@ Tests verify that malicious sparse tensors are rejected before they can trigger
 out-of-bounds memory writes during to_dense() operations.
 """
 
-import base64
 import io
 
+import pybase64 as base64
 import pytest
 import torch