[Hardware][CPU] Add embedding models support for CPU backend (#10193)

Signed-off-by: Isotr0py <2037008807@qq.com>

.buildkite/run-cpu-test-ppc64le.sh

@@ -25,8 +25,7 @@ function cpu_tests() {
       decord einops librosa peft Pillow sentence-transformers soundfile \
       transformers_stream_generator matplotlib datamodel_code_generator
     pip install torchvision --index-url https://download.pytorch.org/whl/cpu
-    # Embedding models are not supported for CPU yet
-    # pytest -v -s tests/models/embedding/language
+    pytest -v -s tests/models/embedding/language
     pytest -v -s tests/models/encoder_decoder/language
     pytest -v -s tests/models/decoder_only/language/test_models.py
     pytest -v -s tests/models/decoder_only/audio_language -m cpu_model

.buildkite/run-cpu-test.sh

@@ -32,8 +32,7 @@ function cpu_tests() {
       decord einops librosa peft Pillow sentence-transformers soundfile \
       transformers_stream_generator matplotlib datamodel_code_generator
     pip install torchvision --index-url https://download.pytorch.org/whl/cpu
-    # Embedding models are not supported for CPU yet
-    # pytest -v -s tests/models/embedding/language
+    pytest -v -s tests/models/embedding/language
     pytest -v -s tests/models/encoder_decoder/language
     pytest -v -s tests/models/decoder_only/language/test_models.py
     pytest -v -s tests/models/decoder_only/audio_language -m cpu_model

tests/models/embedding/language/test_embedding.py

@@ -4,6 +4,8 @@ Run `pytest tests/models/embedding/language/test_embedding.py`.
 """
 import pytest
 
+from vllm.utils import current_platform
+
 from ..utils import check_embeddings_close
 
 # Model, Guard
@@ -21,15 +23,14 @@ ENCODER_ONLY = [
 @pytest.mark.parametrize("model", MODELS)
 @pytest.mark.parametrize("dtype", ["half"])
 def test_models(
-    monkeypatch,
     hf_runner,
     vllm_runner,
     example_prompts,
     model,
     dtype: str,
 ) -> None:
-    if model in ENCODER_ONLY:
-        monkeypatch.setenv("VLLM_ATTENTION_BACKEND", "XFORMERS")
+    if model not in ENCODER_ONLY and current_platform.is_cpu():
+        pytest.skip("Skip large embedding models test on CPU.")
 
     # The example_prompts has ending "\n", for example:
     # "Write a short story about a robot that dreams for the first time.\n"

vllm/attention/backends/torch_sdpa.py

@@ -158,7 +158,8 @@ class TorchSDPAMetadata(AttentionMetadata, PagedAttentionMetadata):
         * Appropriate sequence lengths tensor for key & value
         '''
 
-        if attn_type == AttentionType.DECODER:
+        if (attn_type == AttentionType.DECODER
+                or attn_type == AttentionType.ENCODER_ONLY):
             seq_lens_q = self.seq_lens
             seq_lens_kv = self.seq_lens
         elif attn_type == AttentionType.ENCODER:
@@ -189,7 +190,8 @@ class TorchSDPAMetadata(AttentionMetadata, PagedAttentionMetadata):
         * Appropriate attention bias value given the attention type
         '''
 
-        if attn_type == AttentionType.DECODER:
+        if (attn_type == AttentionType.DECODER
+                or attn_type == AttentionType.ENCODER_ONLY):
             return self.attn_bias
         elif attn_type == AttentionType.ENCODER:
             return self.encoder_attn_bias
@@ -215,7 +217,8 @@ class TorchSDPAMetadata(AttentionMetadata, PagedAttentionMetadata):
                     encoder/decoder cross-attention
         '''
 
-        if attn_type == AttentionType.DECODER:
+        if (attn_type == AttentionType.DECODER
+                or attn_type == AttentionType.ENCODER_ONLY):
             self.attn_bias = attn_bias
         elif attn_type == AttentionType.ENCODER:
             self.encoder_attn_bias = attn_bias
@@ -252,7 +255,8 @@ class TorchSDPAMetadata(AttentionMetadata, PagedAttentionMetadata):
         * Appropriate block tables (or None)
         '''
 
-        if attn_type == AttentionType.DECODER:
+        if (attn_type == AttentionType.DECODER
+                or attn_type == AttentionType.ENCODER_ONLY):
             # Decoder self-attention
             # Choose max_seq_len based on whether we are in prompt_run
             return (self.seq_lens_tensor, self.max_decode_seq_len,
@@ -420,6 +424,8 @@ class TorchSDPABackendImpl(AttentionImpl[TorchSDPAMetadata]):
                     "Torch SDPA backend doesn't support prefix decoding.")
 
         if decode_meta := attn_metadata.decode_metadata:
+            assert attn_type != AttentionType.ENCODER_ONLY, (
+                "Encoder-only models should not have decode metadata.")
             # Decoding run.
             (
                 seq_lens_arg,

vllm/model_executor/models/bert.py

@@ -5,7 +5,6 @@ from torch import nn
 from transformers import BertConfig
 
 from vllm.attention import Attention, AttentionMetadata, AttentionType
-from vllm.attention.backends.xformers import XFormersImpl
 from vllm.config import CacheConfig, VllmConfig
 from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import get_act_fn
@@ -218,11 +217,6 @@ class BertSelfAttention(nn.Module):
                               quant_config=quant_config,
                               prefix=f"{prefix}.attn")
 
-        if not isinstance(self.attn.impl, XFormersImpl):
-            raise ValueError(
-                "Encoder-only models currently require XFORMERS attention "
-                "backend. Set VLLM_ATTENTION_BACKEND=XFORMERS to use BERT.")
-
     def forward(
         self,
         hidden_states: torch.Tensor,

vllm/worker/cpu_embedding_model_runner.py

+import dataclasses
+from typing import Any, Dict, List, Optional, Tuple, Type, Union
+
+import torch
+
+from vllm.model_executor.pooling_metadata import PoolingMetadata
+from vllm.multimodal import MultiModalKwargs
+from vllm.pooling_params import PoolingParams
+from vllm.sequence import (IntermediateTensors, PoolerOutput, SequenceData,
+                           SequenceGroupMetadata)
+from vllm.worker.cpu_model_runner import (CPUModelRunnerBase, ModelInputForCPU,
+                                          ModelInputForCPUBuilder)
+
+
+@dataclasses.dataclass(frozen=True)
+class ModelInputForCPUWithPoolingMetadata(ModelInputForCPU):
+    """
+    Used by the CPUEmbeddingModelRunner.
+    """
+    pooling_metadata: Optional["PoolingMetadata"] = None
+
+
+class CPUEmbeddingModelRunner(
+        CPUModelRunnerBase[ModelInputForCPUWithPoolingMetadata]):
+    _model_input_cls: Type[ModelInputForCPUWithPoolingMetadata] = (
+        ModelInputForCPUWithPoolingMetadata)
+    _builder_cls: Type[ModelInputForCPUBuilder] = ModelInputForCPUBuilder
+
+    @torch.inference_mode()
+    def execute_model(
+        self,
+        model_input: ModelInputForCPUWithPoolingMetadata,
+        kv_caches: List[torch.Tensor],
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+        num_steps: int = 1,
+    ) -> Optional[Union[List[PoolerOutput], IntermediateTensors]]:
+        if num_steps > 1:
+            raise ValueError(
+                "CPU worker does not support multi-step execution.")
+
+        num_layers = self.model_config.get_num_layers(self.parallel_config)
+        # use an empty tensor instead of `None`` to force Dynamo to pass
+        # it by reference, rather by specializing on the value ``None``.
+        # the `dtype` argument does not matter, and we use `float32` as
+        # a placeholder (it has wide hardware support).
+        kv_caches = [
+            torch.tensor([], dtype=torch.float32, device=self.device)
+            for _ in range(num_layers)
+        ]
+
+        model_executable = self.model
+        execute_model_kwargs = {
+            "input_ids":
+            model_input.input_tokens,
+            "positions":
+            model_input.input_positions,
+            "kv_caches":
+            kv_caches,
+            "attn_metadata":
+            model_input.attn_metadata,
+            **MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs or {},
+                                         device=self.device),
+            "intermediate_tensors":
+            intermediate_tensors,
+        }
+
+        hidden_states = model_executable(**execute_model_kwargs)
+
+        return [
+            self.model.pooler(hidden_states=hidden_states,
+                              pooling_metadata=model_input.pooling_metadata)
+        ]
+
+    def make_model_input_from_broadcasted_tensor_dict(
+            self,
+            tensor_dict: Dict[str,
+                              Any]) -> ModelInputForCPUWithPoolingMetadata:
+        return ModelInputForCPUWithPoolingMetadata.from_broadcasted_tensor_dict(
+            tensor_dict,
+            attn_backend=self.attn_backend,
+        )
+
+    def prepare_model_input(
+        self,
+        seq_group_metadata_list: Optional[List[SequenceGroupMetadata]],
+        virtual_engine: int = 0,
+        finished_requests_ids: Optional[List[str]] = None
+    ) -> ModelInputForCPUWithPoolingMetadata:
+        assert seq_group_metadata_list is not None
+        model_input = self._prepare_model_input_tensors(
+            seq_group_metadata_list, finished_requests_ids)
+        # Prepare PoolingMetadata.
+        assert model_input.seq_lens is not None
+        pooling_metadata = self._prepare_pooling(seq_group_metadata_list,
+                                                 model_input.seq_lens)
+
+        return dataclasses.replace(model_input,
+                                   pooling_metadata=pooling_metadata)
+
+    def _prepare_pooling(
+        self,
+        seq_group_metadata_list: List[SequenceGroupMetadata],
+        prompt_lens: List[int],
+    ) -> PoolingMetadata:
+        """Prepare PoolingMetadata for the sequence group metadata list."""
+        seq_groups: List[Tuple[List[int], PoolingParams]] = []
+        for i, seq_group_metadata in enumerate(seq_group_metadata_list):
+            seq_ids = list(seq_group_metadata.seq_data.keys())
+            pooling_params = seq_group_metadata.pooling_params
+            seq_groups.append((seq_ids, pooling_params))
+
+        seq_data: Dict[int, SequenceData] = {}
+        for seq_group_metadata in seq_group_metadata_list:
+            seq_data.update(seq_group_metadata.seq_data)
+
+        pooling_metadata = PoolingMetadata(
+            seq_groups=seq_groups,
+            seq_data=seq_data,
+            prompt_lens=prompt_lens,
+        )
+
+        return pooling_metadata

vllm/worker/cpu_enc_dec_model_runner.py

@@ -8,7 +8,7 @@ from vllm.model_executor.layers.sampler import SamplerOutput
 from vllm.multimodal import MultiModalKwargs
 from vllm.sequence import IntermediateTensors, SequenceGroupMetadata
 from vllm.utils import make_tensor_with_pad
-from vllm.worker.cpu_model_runner import (CPUModelRunner,
+from vllm.worker.cpu_model_runner import (CPUModelRunnerBase,
                                           ModelInputForCPUBuilder,
                                           ModelInputForCPUWithSamplingMetadata)
 from vllm.worker.model_runner_base import (
@@ -50,7 +50,8 @@ class EncoderDecoderModelInputForCPU(ModelInputForCPUWithSamplingMetadata):
             super().from_broadcasted_tensor_dict(tensor_dict, attn_backend))
 
 
-class CPUEncoderDecoderModelRunner(CPUModelRunner):
+class CPUEncoderDecoderModelRunner(
+        CPUModelRunnerBase[EncoderDecoderModelInputForCPU]):
     _model_input_cls: Type[EncoderDecoderModelInputForCPU] = (
         EncoderDecoderModelInputForCPU)
     _builder_cls: Type[ModelInputForCPUBuilder] = ModelInputForCPUBuilder
@@ -87,10 +88,8 @@ class CPUEncoderDecoderModelRunner(CPUModelRunner):
         virtual_engine: int = 0,
         finished_requests_ids: Optional[List[str]] = None
     ) -> EncoderDecoderModelInputForCPU:
-        model_input = super().prepare_model_input(seq_group_metadata_list,
-                                                  virtual_engine,
-                                                  finished_requests_ids)
-        model_input = cast(EncoderDecoderModelInputForCPU, model_input)
+        model_input = self._prepare_model_input_tensors(
+            seq_group_metadata_list, finished_requests_ids)
         (
             attn_metadata,
             encoder_input_tokens_tensor,

vllm/worker/cpu_model_runner.py

@@ -2,7 +2,8 @@ import dataclasses
 import weakref
 from collections import defaultdict
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Type, Union
+from typing import (TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Type,
+                    TypeVar, Union)
 
 import torch
 from torch import nn
@@ -31,6 +32,7 @@ if TYPE_CHECKING:
 
 logger = init_logger(__name__)
 
+TModelInputForCPU = TypeVar('TModelInputForCPU', bound="ModelInputForCPU")
 _PAD_SLOT_ID = -1
 
 
@@ -60,10 +62,10 @@ class ModelInputForCPU(ModelRunnerInputBase):
 
     @classmethod
     def from_broadcasted_tensor_dict(
-        cls: Type["ModelInputForCPU"],
+        cls: Type[TModelInputForCPU],
         tensor_dict: Dict[str, Any],
         attn_backend: Optional["AttentionBackend"] = None
-    ) -> "ModelInputForCPU":
+    ) -> TModelInputForCPU:
         if attn_backend is not None:
             tensor_dict = _init_attn_metadata_from_tensor_dict(
                 attn_backend, tensor_dict)
@@ -255,6 +257,9 @@ class ModelInputForCPUBuilder(ModelRunnerInputBuilderBase[ModelInputForCPU]):
                     slot_mapping.append(_PAD_SLOT_ID)
                     continue
 
+                # For encoder-only models, the block_table is None,
+                # and there is no need to initialize the slot_mapping.
+                if block_table is not None:
                     block_number = block_table[i //
                                                self.block_size]  # type: ignore
                     block_offset = i % self.block_size  # type: ignore
@@ -402,10 +407,12 @@ class ModelInputForCPUBuilder(ModelRunnerInputBuilderBase[ModelInputForCPU]):
         )
 
 
-class CPUModelRunner(ModelRunnerBase[ModelInputForCPU]):
-    _model_input_cls: Type[ModelInputForCPUWithSamplingMetadata] = (
-        ModelInputForCPUWithSamplingMetadata)
-    _builder_cls: Type[ModelInputForCPUBuilder] = ModelInputForCPUBuilder
+class CPUModelRunnerBase(ModelRunnerBase[TModelInputForCPU]):
+    """
+    Helper class for shared methods between CPU model runners.
+    """
+    _model_input_cls: Type[TModelInputForCPU]
+    _builder_cls: Type[ModelInputForCPUBuilder]
 
     def __init__(
         self,
@@ -448,20 +455,11 @@ class CPUModelRunner(ModelRunnerBase[ModelInputForCPU]):
     def load_model(self) -> None:
         self.model = get_model(vllm_config=self.vllm_config)
 
-    def make_model_input_from_broadcasted_tensor_dict(
-        self,
-        tensor_dict: Dict[str, Any],
-    ) -> ModelInputForCPUWithSamplingMetadata:
-        return ModelInputForCPUWithSamplingMetadata.from_broadcasted_tensor_dict(  # noqa: E501
-            tensor_dict,
-            attn_backend=self.attn_backend,
-        )
-
     def _prepare_model_input_tensors(
         self,
         seq_group_metadata_list: List[SequenceGroupMetadata],
         finished_requests_ids: Optional[List[str]] = None
-    ) -> ModelInputForCPUWithSamplingMetadata:
+    ) -> TModelInputForCPU:
         """Helper method to prepare the model input based on a given sequence
         group. Prepares metadata needed for the base model forward pass but not
         metadata for possible additional steps, e.g., sampling.
@@ -473,6 +471,21 @@ class CPUModelRunner(ModelRunnerBase[ModelInputForCPU]):
 
         return builder.build()  # type: ignore
 
+
+class CPUModelRunner(CPUModelRunnerBase[ModelInputForCPUWithSamplingMetadata]):
+    _model_input_cls: Type[ModelInputForCPUWithSamplingMetadata] = (
+        ModelInputForCPUWithSamplingMetadata)
+    _builder_cls: Type[ModelInputForCPUBuilder] = ModelInputForCPUBuilder
+
+    def make_model_input_from_broadcasted_tensor_dict(
+        self,
+        tensor_dict: Dict[str, Any],
+    ) -> ModelInputForCPUWithSamplingMetadata:
+        return ModelInputForCPUWithSamplingMetadata.from_broadcasted_tensor_dict(  # noqa: E501
+            tensor_dict,
+            attn_backend=self.attn_backend,
+        )
+
     def prepare_model_input(
         self,
         seq_group_metadata_list: List[SequenceGroupMetadata],

vllm/worker/cpu_worker.py

@@ -14,8 +14,9 @@ from vllm.logger import init_logger
 from vllm.model_executor import set_random_seed
 from vllm.sequence import ExecuteModelRequest
 from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE
+from vllm.worker.cpu_embedding_model_runner import CPUEmbeddingModelRunner
 from vllm.worker.cpu_enc_dec_model_runner import CPUEncoderDecoderModelRunner
-from vllm.worker.cpu_model_runner import CPUModelRunner
+from vllm.worker.cpu_model_runner import CPUModelRunner, CPUModelRunnerBase
 from vllm.worker.worker_base import (LocalOrDistributedWorkerBase,
                                      LoraNotSupportedWorkerBase, WorkerBase,
                                      WorkerInput)
@@ -150,21 +151,20 @@ class CPUWorker(LoraNotSupportedWorkerBase, LocalOrDistributedWorkerBase):
         else:
             self.local_omp_cpuid = omp_cpuids.split("|")[rank]
 
-        ModelRunnerClass: Type[CPUModelRunner] = CPUModelRunner
+        ModelRunnerClass: Type[CPUModelRunnerBase] = CPUModelRunner
         if self.model_config.task == "embedding":
-            raise NotImplementedError(
-                "Embedding models are not supported for CPU backend")
-            # ModelRunnerClass = CPUEmbeddingModelRunner
+            ModelRunnerClass = CPUEmbeddingModelRunner
         elif self.model_config.is_encoder_decoder:
             ModelRunnerClass = CPUEncoderDecoderModelRunner
-        self.model_runner: CPUModelRunner = ModelRunnerClass(
+        self.model_runner: CPUModelRunnerBase = ModelRunnerClass(
             vllm_config=vllm_config,
             kv_cache_dtype=kv_cache_dtype,
             is_driver_worker=is_driver_worker)
         # Uninitialized cache engine. Will be initialized by
         # initialize_cache.
         self.cache_engine: List[CPUCacheEngine]
-        self.cpu_cache: List[List[torch.Tensor]]
+        # Initialize cpu_cache as embedding models don't initialize kv_caches
+        self.cpu_cache: Optional[List[List[torch.Tensor]]] = None
 
         # Torch profiler. Enabled and configured through env vars:
         # VLLM_TORCH_PROFILER_DIR=/path/to/save/trace