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
@@ -1,7 +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.
|
||||
|
||||
Reference in New Issue
Block a user