diff --git a/docs/models/pooling_models/embed.md b/docs/models/pooling_models/embed.md index 4c33126977..5211997edb 100644 --- a/docs/models/pooling_models/embed.md +++ b/docs/models/pooling_models/embed.md @@ -37,7 +37,7 @@ You can compute pairwise similarity scores to build a similarity matrix using th | Architecture | Models | Example HF Models | [LoRA](../../features/lora.md) | [PP](../../serving/parallelism_scaling.md) | | ------------ | ------ | ----------------- | ------------------------------ | ------------------------------------------ | -| `BertModel` | BERT-based | `BAAI/bge-base-en-v1.5`, `Snowflake/snowflake-arctic-embed-xs`, etc. | | | +| `BertModel` | BERT-based | `BAAI/bge-base-en-v1.5`, `Snowflake/snowflake-arctic-embed-xs`, etc. | ✅︎ | | | `BertSpladeSparseEmbeddingModel` | SPLADE | `naver/splade-v3` | | | | `BgeM3EmbeddingModel` | BGE-M3 | `BAAI/bge-m3` | | | | `Gemma2Model`C | Gemma 2-based | `BAAI/bge-multilingual-gemma2`, etc. | ✅︎ | ✅︎ | @@ -51,9 +51,9 @@ You can compute pairwise similarity scores to build a similarity matrix using th | `NomicBertModel` | Nomic BERT | `nomic-ai/nomic-embed-text-v1`, `nomic-ai/nomic-embed-text-v2-moe`, `Snowflake/snowflake-arctic-embed-m-long`, etc. | | | | `Qwen2Model`C, `Qwen2ForCausalLM`C | Qwen2-based | `ssmits/Qwen2-7B-Instruct-embed-base` (see note), `Alibaba-NLP/gte-Qwen2-7B-instruct` (see note), etc. | ✅︎ | ✅︎ | | `Qwen3Model`C, `Qwen3ForCausalLM`C | Qwen3-based | `Qwen/Qwen3-Embedding-0.6B`, `ai-sage/Giga-Embeddings-instruct-480M-0826`^, etc. | ✅︎ | ✅︎ | -| `RobertaModel`, `RobertaForMaskedLM` | RoBERTa-based | `sentence-transformers/all-roberta-large-v1`, etc. | | | +| `RobertaModel`, `RobertaForMaskedLM` | RoBERTa-based | `sentence-transformers/all-roberta-large-v1`, etc. | ✅︎ | | | `VoyageQwen3BidirectionalEmbedModel`C | Voyage Qwen3-based with bidirectional attention | `voyageai/voyage-4-nano`, etc. | ✅︎ | ✅︎ | -| `XLMRobertaModel` | XLMRobertaModel-based | `BAAI/bge-m3` (see note), `intfloat/multilingual-e5-base`, `jinaai/jina-embeddings-v3` (see note), etc. | | | +| `XLMRobertaModel` | XLMRobertaModel-based | `BAAI/bge-m3` (see note), `intfloat/multilingual-e5-base`, `jinaai/jina-embeddings-v3` (see note), etc. | ✅︎ | | | `*Model`C, `*ForCausalLM`C, etc. | Generative models | N/A | \* | \* | !!! note diff --git a/vllm/model_executor/models/bert.py b/vllm/model_executor/models/bert.py index cdaa0bce97..5087486812 100644 --- a/vllm/model_executor/models/bert.py +++ b/vllm/model_executor/models/bert.py @@ -41,7 +41,7 @@ from vllm.sequence import IntermediateTensors from vllm.tasks import PoolingTask from vllm.v1.pool.metadata import PoolingMetadata -from .interfaces import SupportsCrossEncoding, SupportsQuant +from .interfaces import SupportsCrossEncoding, SupportsLoRA, SupportsQuant from .interfaces_base import attn_type, default_pooling_type from .utils import AutoWeightsLoader, WeightsMapper, maybe_prefix @@ -448,7 +448,7 @@ class BertPoolingModel(BertModel): @default_pooling_type(seq_pooling_type="CLS") -class BertEmbeddingModel(nn.Module, SupportsQuant): +class BertEmbeddingModel(nn.Module, SupportsQuant, SupportsLoRA): """A model that uses Bert to provide embedding functionalities. This class encapsulates the BertModel and provides an interface for @@ -463,9 +463,22 @@ class BertEmbeddingModel(nn.Module, SupportsQuant): is_pooling_model = True embedding_class: type[nn.Module] = BertEmbedding + packed_modules_mapping = {"qkv_proj": ["query", "key", "value"]} + # PEFT adapters name modules relative to the HF backbone (optionally under + # `bert.`/`roberta.`), while vLLM nests the backbone under `model.`. + hf_to_vllm_mapper = WeightsMapper( + orig_to_new_prefix={ + "bert.": "model.", + "roberta.": "model.", + "encoder.": "model.encoder.", + "embeddings.": "model.embeddings.", + } + ) + def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""): super().__init__() + self.config = vllm_config.model_config.hf_config pooler_config = vllm_config.model_config.pooler_config assert pooler_config is not None