0
# Text Embeddings
1

2
Text encoding and embedding generation for semantic similarity, retrieval applications, and downstream NLP tasks. Supports various pooling strategies and normalization options for different use cases.
3

4
## Capabilities
5

6
### Text Encoding
7

8
Generate dense vector representations of text for semantic tasks, similarity search, and downstream machine learning applications.
9

10
```python { .api }
11
def encode(
12
    self,
13
    prompts: Union[PromptType, Sequence[PromptType], DataPrompt],
14
    pooling_params: Optional[Union[PoolingParams, Sequence[PoolingParams]]] = None,
15
    *,
16
    truncate_prompt_tokens: Optional[int] = None,
17
    use_tqdm: Union[bool, Callable[..., tqdm]] = True,
18
    lora_request: Optional[Union[List[LoRARequest], LoRARequest]] = None,
19
    pooling_task: PoolingTask = "encode",
20
    tokenization_kwargs: Optional[Dict[str, Any]] = None
21
) -> List[PoolingRequestOutput]:
22
    """
23
    Generate embeddings for input text.
24

25
    Parameters:
26
    - prompts: Input text or token sequences
27
    - pooling_params: Pooling strategy and normalization options
28
    - truncate_prompt_tokens: Maximum prompt length (keyword-only)
29
    - use_tqdm: Show progress bar (keyword-only)
30
    - lora_request: LoRA adapter configuration (keyword-only)
31
    - pooling_task: The pooling task to perform (keyword-only)
32
    - tokenization_kwargs: Additional tokenization options (keyword-only)
33

34
    Returns:
35
    List of PoolingRequestOutput with vector representations
36
    """
37
```
38

39
## Usage Examples
40

41
### Basic Text Embeddings
42

43
```python
44
from vllm import LLM, PoolingParams
45

46
llm = LLM(model="sentence-transformers/all-MiniLM-L6-v2")
47

48
texts = [
49
    "The quick brown fox jumps over the lazy dog.",
50
    "A fast fox leaps over a sleeping dog.",
51
    "Python is a programming language."
52
]
53

54
pooling_params = PoolingParams(pooling_type="MEAN", normalize=True)
55
outputs = llm.encode(texts, pooling_params=pooling_params)
56

57
for output in outputs:
58
    print(f"Embedding dimension: {len(output.outputs.data)}")
59
```
60

61
## Types
62

63
```python { .api }
64
class PoolingRequestOutput:
65
    id: str
66
    outputs: PoolingOutput
67
    prompt_token_ids: List[int]
68
    finished: bool
69

70
class PoolingOutput:
71
    data: List[float]  # Dense vector representation
72
```