Add pydocstyle to the ruff rules (#52136)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Signed-off-by: Harry Mellor <hej.mellor@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
Example script demonstrating long text embedding with chunked processing in vLLM.
|
||||
"""Example script demonstrating long text embedding with chunked processing in vLLM.
|
||||
|
||||
This example shows how to use vLLM's chunked processing feature to handle text
|
||||
inputs that exceed the model's maximum token length. The feature automatically
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
"""
|
||||
This example shows how to use vLLM for running offline inference with
|
||||
"""This example shows how to use vLLM for running offline inference with
|
||||
the correct prompt format on vision language models for multimodal embedding.
|
||||
|
||||
For most models, the prompt format should follow corresponding examples
|
||||
|
||||
@@ -37,8 +37,7 @@ def create_chat_embeddings(
|
||||
add_special_tokens: bool = False,
|
||||
padding: str | NotGiven = NOT_GIVEN,
|
||||
) -> CreateEmbeddingResponse:
|
||||
"""
|
||||
Convenience function for accessing vLLM's Chat Embeddings API,
|
||||
"""Convenience function for accessing vLLM's Chat Embeddings API,
|
||||
which is an extension of OpenAI's existing Embeddings API.
|
||||
"""
|
||||
return client.post(
|
||||
@@ -56,13 +55,11 @@ def create_chat_embeddings(
|
||||
|
||||
|
||||
def run_clip(client: OpenAI, model: str):
|
||||
"""
|
||||
Start the server using:
|
||||
"""Start the server using:
|
||||
|
||||
vllm serve openai/clip-vit-base-patch32 \
|
||||
--runner pooling
|
||||
"""
|
||||
|
||||
response = create_chat_embeddings(
|
||||
client,
|
||||
messages=[
|
||||
@@ -97,8 +94,7 @@ def run_clip(client: OpenAI, model: str):
|
||||
|
||||
|
||||
def run_dse_qwen2_vl(client: OpenAI, model: str):
|
||||
"""
|
||||
Start the server using:
|
||||
"""Start the server using:
|
||||
|
||||
vllm serve MrLight/dse-qwen2-2b-mrl-v1 \
|
||||
--runner pooling \
|
||||
@@ -159,14 +155,12 @@ def run_dse_qwen2_vl(client: OpenAI, model: str):
|
||||
|
||||
|
||||
def run_qwen3_vl(client: OpenAI, model: str):
|
||||
"""
|
||||
Start the server using:
|
||||
"""Start the server using:
|
||||
|
||||
vllm serve Qwen/Qwen3-VL-Embedding-2B \
|
||||
--runner pooling \
|
||||
--max-model-len 8192
|
||||
"""
|
||||
|
||||
default_instruction = "Represent the user's input."
|
||||
|
||||
print("Text embedding output:")
|
||||
@@ -266,14 +260,12 @@ def run_qwen3_vl(client: OpenAI, model: str):
|
||||
|
||||
|
||||
def run_siglip(client: OpenAI, model: str):
|
||||
"""
|
||||
Start the server using:
|
||||
"""Start the server using:
|
||||
|
||||
vllm serve google/siglip-base-patch16-224 \
|
||||
--runner pooling \
|
||||
--chat-template template_basic.jinja
|
||||
"""
|
||||
|
||||
response = create_chat_embeddings(
|
||||
client,
|
||||
messages=[
|
||||
@@ -311,8 +303,7 @@ def run_siglip(client: OpenAI, model: str):
|
||||
|
||||
|
||||
def run_vlm2vec(client: OpenAI, model: str):
|
||||
"""
|
||||
Start the server using:
|
||||
"""Start the server using:
|
||||
|
||||
vllm serve TIGER-Lab/VLM2Vec-Full \
|
||||
--runner pooling \
|
||||
@@ -320,7 +311,6 @@ def run_vlm2vec(client: OpenAI, model: str):
|
||||
--max-model-len 4096 \
|
||||
--chat-template examples/pooling/embed/template/vlm2vec_phi3v.jinja
|
||||
"""
|
||||
|
||||
response = create_chat_embeddings(
|
||||
client,
|
||||
messages=[
|
||||
|
||||
@@ -85,8 +85,7 @@ def generate_datamodule():
|
||||
|
||||
|
||||
def process_channel_group(orig_img, channels):
|
||||
"""
|
||||
Args:
|
||||
"""Args:
|
||||
orig_img: torch.Tensor representing original image (reference)
|
||||
with shape = (bands, H, W).
|
||||
channels: list of indices representing RGB channels.
|
||||
@@ -94,8 +93,8 @@ def process_channel_group(orig_img, channels):
|
||||
Returns:
|
||||
torch.Tensor with shape (num_channels, height, width)
|
||||
for original image
|
||||
"""
|
||||
|
||||
"""
|
||||
orig_img = orig_img[channels, ...]
|
||||
valid_mask = torch.ones_like(orig_img, dtype=torch.bool)
|
||||
valid_mask[orig_img == NO_DATA_FLOAT] = False
|
||||
@@ -121,8 +120,8 @@ def read_geotiff(file_path: str):
|
||||
Returns:
|
||||
np.ndarray with shape (bands, height, width)
|
||||
meta info dict
|
||||
"""
|
||||
|
||||
"""
|
||||
with rasterio.open(file_path) as src:
|
||||
img = src.read()
|
||||
meta = src.meta
|
||||
@@ -142,8 +141,8 @@ def save_geotiff(image, output_path: str, meta: dict):
|
||||
image: np.ndarray with shape (bands, height, width)
|
||||
output_path: path where to save the image
|
||||
meta: dict with meta info.
|
||||
"""
|
||||
|
||||
"""
|
||||
with rasterio.open(output_path, "w", **meta) as dest:
|
||||
for i in range(image.shape[0]):
|
||||
dest.write(image[i, :, :], i + 1)
|
||||
@@ -172,12 +171,14 @@ def load_example(
|
||||
images in *file_paths*.
|
||||
std: list containing std values for each band in the
|
||||
images in *file_paths*.
|
||||
indices: optional band indices to select from each image. All bands
|
||||
are kept when None.
|
||||
|
||||
Returns:
|
||||
np.array containing created example
|
||||
list of meta info for each image in *file_paths*
|
||||
"""
|
||||
|
||||
"""
|
||||
imgs = []
|
||||
metas = []
|
||||
temporal_coords = []
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
Example offline usage of sequence reward models.
|
||||
"""Example offline usage of sequence reward models.
|
||||
|
||||
The key distinction between sequence classification and token classification
|
||||
lies in their output granularity: sequence classification produces a single
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example online usage of sequence reward models.
|
||||
"""Example online usage of sequence reward models.
|
||||
|
||||
Run `vllm serve <model> --runner pooling`
|
||||
to start up the server in vLLM. e.g.
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
Example offline usage of token reward models.
|
||||
"""Example offline usage of token reward models.
|
||||
|
||||
The key distinction between sequence classification and token classification
|
||||
lies in their output granularity: sequence classification produces a single
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example online usage of token reward models.
|
||||
"""Example online usage of token reward models.
|
||||
|
||||
Run `vllm serve <model> --runner pooling`
|
||||
to start up the server in vLLM. e.g.
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example of using the OpenAI entrypoint's rerank API which is compatible with
|
||||
"""Example of using the OpenAI entrypoint's rerank API which is compatible with
|
||||
the Cohere SDK: https://github.com/cohere-ai/cohere-python
|
||||
Note that `pip install cohere` is needed to run this example.
|
||||
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example of using ColBERT late interaction models for reranking and scoring.
|
||||
"""Example of using ColBERT late interaction models for reranking and scoring.
|
||||
|
||||
ColBERT (Contextualized Late Interaction over BERT) uses per-token embeddings
|
||||
and MaxSim scoring for document reranking, providing better accuracy than
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example of using ColModernVBERT late interaction model for reranking.
|
||||
"""Example of using ColModernVBERT late interaction model for reranking.
|
||||
|
||||
ColModernVBERT is a multi-modal ColBERT-style model combining a SigLIP
|
||||
vision encoder with a ModernBERT text encoder. It produces per-token
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example of using ColQwen3.5 late interaction model for reranking.
|
||||
"""Example of using ColQwen3.5 late interaction model for reranking.
|
||||
|
||||
ColQwen3.5 is a multi-modal ColBERT-style model based on Qwen3.5.
|
||||
It produces per-token embeddings and uses MaxSim scoring for retrieval
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
"""
|
||||
Example of using ColQwen3 late interaction model for reranking and scoring.
|
||||
"""Example of using ColQwen3 late interaction model for reranking and scoring.
|
||||
|
||||
ColQwen3 is a multi-modal ColBERT-style model based on Qwen3-VL.
|
||||
It produces per-token embeddings and uses MaxSim scoring for retrieval
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
Script to convert Large Language Models (LLMs) to Sequence Classification models.
|
||||
"""Script to convert Large Language Models (LLMs) to Sequence Classification models.
|
||||
This is particularly useful for converting reranker models that use next-token
|
||||
prediction to a sequence classification format for compatibility with standard
|
||||
classification and rerank pipelines.
|
||||
@@ -35,8 +34,7 @@ import transformers
|
||||
|
||||
|
||||
def from_2_way_softmax(causal_lm, seq_cls_model, tokenizer, tokens, device):
|
||||
"""
|
||||
This method extracts the difference between weights for 'true' and 'false' tokens
|
||||
"""This method extracts the difference between weights for 'true' and 'false' tokens
|
||||
from the language model head to create a single classification weight vector.
|
||||
|
||||
Args:
|
||||
@@ -47,6 +45,7 @@ def from_2_way_softmax(causal_lm, seq_cls_model, tokenizer, tokens, device):
|
||||
device: Target device (cpu/cuda)
|
||||
|
||||
Reference: https://huggingface.co/Qwen/Qwen3-Reranker-0.6B/discussions/3
|
||||
|
||||
"""
|
||||
assert len(tokens) == 2, (
|
||||
"Method requires exactly two tokens for binary classification"
|
||||
@@ -73,8 +72,7 @@ def from_2_way_softmax(causal_lm, seq_cls_model, tokenizer, tokens, device):
|
||||
|
||||
|
||||
def no_post_processing(causal_lm, seq_cls_model, tokenizer, tokens, device):
|
||||
"""
|
||||
Directly use token weights from the language model head for classification.
|
||||
"""Directly use token weights from the language model head for classification.
|
||||
|
||||
This method maps each classification label directly to a corresponding token
|
||||
in the vocabulary without additional transformation.
|
||||
@@ -85,6 +83,7 @@ def no_post_processing(causal_lm, seq_cls_model, tokenizer, tokens, device):
|
||||
tokenizer: Model tokenizer
|
||||
tokens: List of tokens representing class labels
|
||||
device: Target device (cpu/cuda)
|
||||
|
||||
"""
|
||||
# Get the language model head weights (vocabulary_size x hidden_size)
|
||||
lm_head_weights = causal_lm.lm_head.weight
|
||||
@@ -110,8 +109,7 @@ method_map = {
|
||||
def converting(
|
||||
model_name, classifier_from_tokens, path, method, use_sep_token=False, device="cpu"
|
||||
):
|
||||
"""
|
||||
Main conversion function to transform a CausalLM model to SequenceClassification.
|
||||
"""Main conversion function to transform a CausalLM model to SequenceClassification.
|
||||
|
||||
Args:
|
||||
model_name: Name or path of the pretrained model
|
||||
@@ -120,6 +118,7 @@ def converting(
|
||||
method: Conversion method ('from_2_way_softmax' or 'no_post_processing')
|
||||
use_sep_token: Whether to use separating token in the sequence classification model
|
||||
device: Device to load the model on ('cpu' or 'cuda')
|
||||
|
||||
"""
|
||||
assert method in method_map, f"Unknown method: {method}"
|
||||
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
What is the difference between the official original version and one
|
||||
"""What is the difference between the official original version and one
|
||||
that has been converted into a sequence classification model?
|
||||
|
||||
Qwen3-Reranker is a language model that doing reranker by using the
|
||||
@@ -34,8 +33,7 @@ model_name = "Qwen/Qwen3-Reranker-0.6B"
|
||||
|
||||
|
||||
def get_llm() -> LLM:
|
||||
"""
|
||||
Initializes and returns the LLM model for Qwen3-Reranker.
|
||||
"""Initializes and returns the LLM model for Qwen3-Reranker.
|
||||
|
||||
Returns:
|
||||
LLM: Configured vLLM instance for reranking tasks.
|
||||
@@ -43,6 +41,7 @@ def get_llm() -> LLM:
|
||||
Note:
|
||||
This function loads the ORIGINAL Qwen3-Reranker model with specific
|
||||
overrides to make it compatible with vLLM's score API.
|
||||
|
||||
"""
|
||||
return LLM(
|
||||
# Specify the original model from HuggingFace
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
"""
|
||||
What is the difference between the official original version and one
|
||||
"""What is the difference between the official original version and one
|
||||
that has been converted into a sequence classification model?
|
||||
|
||||
Qwen3-Reranker is a language model that doing reranker by using the
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example of using the OpenAI entrypoint's rerank API which is compatible with
|
||||
"""Example of using the OpenAI entrypoint's rerank API which is compatible with
|
||||
Jina and Cohere https://jina.ai/reranker
|
||||
|
||||
run: vllm serve BAAI/bge-reranker-base
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Example online usage of Score API.
|
||||
"""Example online usage of Score API.
|
||||
|
||||
Run `vllm serve <model> --runner pooling` to start up the server in vLLM.
|
||||
"""
|
||||
|
||||
@@ -68,13 +68,13 @@ def get_chat_template(model: str) -> str:
|
||||
def get_hf_overrides(model: str) -> dict[str, Any]:
|
||||
"""Convert Large Language Models (LLMs) to Sequence Classification models.
|
||||
|
||||
note:
|
||||
Note:
|
||||
Some reranking models require special configuration overrides to work
|
||||
correctly with vLLM's score API.
|
||||
Reference: https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/qwen3_reranker_offline.py
|
||||
Reference: https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/convert_model_to_seq_cls.py
|
||||
"""
|
||||
|
||||
"""
|
||||
model_name_to_hf_overrides_map = {
|
||||
"BAAI/bge-reranker-v2-gemma": {
|
||||
"architectures": ["GemmaForSequenceClassification"],
|
||||
@@ -122,7 +122,6 @@ def get_hf_overrides(model: str) -> dict[str, Any]:
|
||||
|
||||
def main(args: Namespace):
|
||||
"""Main execution function for the reranking example."""
|
||||
|
||||
# Get the overrides for the specified model
|
||||
args.hf_overrides = get_hf_overrides(args.model)
|
||||
|
||||
|
||||
@@ -1,15 +1,14 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
"""
|
||||
Example of using the rerank API with template.
|
||||
"""Example of using the rerank API with template.
|
||||
|
||||
This script demonstrates how to interact with a vLLM server running
|
||||
a reranking model via the REST API.
|
||||
Before running this script, start the vLLM server with one of the
|
||||
supported reranking models using the commands below.
|
||||
|
||||
note:
|
||||
Note:
|
||||
Some reranking models require special configuration overrides to work correctly
|
||||
with vLLM's score API.
|
||||
Reference: https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/qwen3_reranker_online.py
|
||||
@@ -21,6 +20,7 @@ run:
|
||||
vllm serve mixedbread-ai/mxbai-rerank-base-v2 --hf_overrides '{"architectures": ["Qwen2ForSequenceClassification"],"classifier_from_token": ["0", "1"], "method": "from_2_way_softmax"}' --chat-template examples/pooling/score/template/mxbai_rerank_v2.jinja
|
||||
vllm serve nvidia/llama-nemotron-rerank-1b-v2 --runner pooling --trust-remote-code --chat-template examples/pooling/score/template/nemotron-rerank.jinja
|
||||
vllm serve Qwen/Qwen3-Reranker-0.6B --runner pooling --hf_overrides '{"architectures": ["Qwen3ForSequenceClassification"],"classifier_from_token": ["no", "yes"],"is_original_qwen3_reranker": true}' --chat-template examples/pooling/score/template/qwen3_reranker.jinja
|
||||
|
||||
"""
|
||||
|
||||
import json
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
Example Python client for multimodal rerank API which is compatible with
|
||||
"""Example Python client for multimodal rerank API which is compatible with
|
||||
Jina and Cohere https://jina.ai/reranker
|
||||
|
||||
Run `vllm serve <model> --runner pooling` to start up the server in vLLM.
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
This example shows how to use vLLM for running offline inference with
|
||||
"""This example shows how to use vLLM for running offline inference with
|
||||
vision language reranker models for multimodal scoring tasks.
|
||||
|
||||
Vision language rerankers score the relevance between a text query and
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
Example online usage of Score API.
|
||||
"""Example online usage of Score API.
|
||||
|
||||
Run `vllm serve <model> --runner pooling` to start up the server in vLLM.
|
||||
e.g.
|
||||
|
||||
@@ -3,8 +3,7 @@
|
||||
# Adapted from Qwen3-ForcedAligner inference:
|
||||
# https://github.com/QwenLM/Qwen3-ASR
|
||||
|
||||
"""
|
||||
Offline forced alignment example using Qwen3-ForcedAligner-0.6B.
|
||||
"""Offline forced alignment example using Qwen3-ForcedAligner-0.6B.
|
||||
|
||||
Forced alignment takes audio and reference text as input and produces
|
||||
word-level timestamps. The model predicts a time bin at each <timestamp>
|
||||
|
||||
@@ -3,8 +3,7 @@
|
||||
# Adapted from Qwen3-ForcedAligner inference:
|
||||
# https://github.com/QwenLM/Qwen3-ASR
|
||||
|
||||
"""
|
||||
Online forced alignment example using Qwen3-ForcedAligner-0.6B-hf.
|
||||
"""Online forced alignment example using Qwen3-ForcedAligner-0.6B-hf.
|
||||
|
||||
Forced alignment takes audio and reference text as input and produces
|
||||
word-level timestamps. The model predicts a time bin at each <timestamp>
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# Adapted from https://huggingface.co/boltuix/NeuroBERT-NER
|
||||
|
||||
"""
|
||||
Example online usage of Pooling API for Named Entity Recognition (NER).
|
||||
"""Example online usage of Pooling API for Named Entity Recognition (NER).
|
||||
|
||||
Run `vllm serve <model> --runner pooling`
|
||||
to start up the server in vLLM. e.g.
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
Example online usage of Pooling API for ColQwen3 multi-vector retrieval.
|
||||
"""Example online usage of Pooling API for ColQwen3 multi-vector retrieval.
|
||||
|
||||
ColQwen3 is a multi-modal late interaction model based on Qwen3-VL that
|
||||
produces per-token embeddings (320-dim, L2-normalized) for both text and
|
||||
|
||||
@@ -2,8 +2,7 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
# ruff: noqa: E501
|
||||
|
||||
"""
|
||||
Example online usage of the Jina Reranker v3 score and rerank APIs with a task
|
||||
"""Example online usage of the Jina Reranker v3 score and rerank APIs with a task
|
||||
instruction.
|
||||
|
||||
Run `vllm serve jinaai/jina-reranker-v3 --runner pooling` to start up the
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
Example online usage of Pooling API for multi vector retrieval.
|
||||
"""Example online usage of Pooling API for multi vector retrieval.
|
||||
|
||||
Run `vllm serve <model> --runner pooling`
|
||||
to start up the server in vLLM. e.g.
|
||||
|
||||
Reference in New Issue
Block a user