[Question]: cannot import name 'BM25Retriever' from 'pipelines.nodes' (/usr/local/python3.7.0/lib/python3.7/site-packages/pipelines-0.1.0a0-py3.7.egg/pipelines/nodes/__init__.py)

[PNightOwlY]

请提出你的问题

docker pull registry.baidubce.com/paddlepaddle/paddlenlp:2.4.0-gpu-cuda10.2-cudnn7 nvidia-docker run -d --name paddlenlp_pipelines_gpu --net host -ti registry.baidubce.com/paddlepaddle/paddlenlp:2.4.0-gpu-cuda10.2-cudnn7

安装的gpu镜像，pip list | grep paddle 查看paddle的版本为 paddle-bfloat 0.1.7 paddle2onnx 0.9.8 paddlefsl 1.1.0 paddlenlp 2.3.0.dev0 paddleocr 2.5.0.3 paddlepaddle-gpu 2.3.1

运行多路召回的example 无法找到对应的BM25Retriever node

[w5688414]

您好，多路召回在0.5版本才加入，需要您升级成0.5版本后才可以使用。Docker镜像需要按照教程，用最新的Paddle的Docker重新打一个。 https://github.com/PaddlePaddle/PaddleNLP/tree/develop/pipelines/docker

[PNightOwlY]

您好，多路召回在0.5版本才加入，需要您升级成0.5版本后才可以使用。Docker镜像需要按照教程，用最新的Paddle的Docker重新打一个。 https://github.com/PaddlePaddle/PaddleNLP/tree/develop/pipelines/docker

谢谢回复！我通过下载paddle的安装包，然后把缺失的环境包都替换了，也成功了！

但是我昨天遇到一个效果问题，我在医疗这个数据集上进行了base和nano的测试，发现base的效果要比nano的效果差，请问这是什么原因呢？

base的配置

version: '1.1.0'

components:    # define all the building-blocks for Pipeline
  - name: DocumentStore
    type: ElasticsearchDocumentStore  # consider using MilvusDocumentStore or WeaviateDocumentStore for scaling to large number of documents
    params:
      host: 172.18.159.16
      port: 9200
      index: ccks_base_encoder
      embedding_dim: 768
  - name: Retriever
    type: DensePassageRetriever
    params:
      document_store: DocumentStore    # params can reference other components defined in the YAML
      top_k: 10
      query_embedding_model: rocketqa-zh-base-query-encoder
      passage_embedding_model: rocketqa-zh-base-para-encoder
      embed_title: False
  - name: Ranker       # custom-name for the component; helpful for visualization & debugging
    type: ErnieRanker    # pipelines Class name for the component
    params:
      model_name_or_path: rocketqa-base-cross-encoder
      top_k: 3
  - name: TextFileConverter
    type: TextConverter
  - name: ImageFileConverter
    type: ImageToTextConverter
  - name: PDFFileConverter
    type: PDFToTextConverter
  - name: DocxFileConverter
    type: DocxToTextConverter
  - name: Preprocessor
    type: PreProcessor
    params:
      split_by: word
      split_length: 1000
  - name: FileTypeClassifier
    type: FileTypeClassifier

pipelines:
  - name: query    # a sample extractive-qa Pipeline
    type: Query
    nodes:
      - name: Retriever
        inputs: [Query]
      - name: Ranker
        inputs: [Retriever]
  - name: indexing
    type: Indexing
    nodes:
      - name: FileTypeClassifier
        inputs: [File]
      - name: TextFileConverter
        inputs: [FileTypeClassifier.output_1]
      - name: PDFFileConverter
        inputs: [FileTypeClassifier.output_2]
      - name: DocxFileConverter
        inputs: [FileTypeClassifier.output_4]
      - name: ImageFileConverter
        inputs: [FileTypeClassifier.output_6]
      - name: Preprocessor
        inputs: [PDFFileConverter, TextFileConverter, DocxFileConverter, ImageFileConverter]
      - name: Retriever
        inputs: [Preprocessor]
      - name: DocumentStore
        inputs: [Retriever]

nano

version: '1.1.0'

components:    # define all the building-blocks for Pipeline
  - name: DocumentStore
    type: ElasticsearchDocumentStore  # consider using MilvusDocumentStore or WeaviateDocumentStore for scaling to large number of documents
    params:
      host: 172.18.159.16
      port: 9200
      index: ccks_encoder
      embedding_dim: 312
  - name: Retriever
    type: DensePassageRetriever
    params:
      document_store: DocumentStore    # params can reference other components defined in the YAML
      top_k: 10
      query_embedding_model: rocketqa-zh-nano-query-encoder
      passage_embedding_model: rocketqa-zh-nano-para-encoder
      embed_title: False
  - name: Ranker       # custom-name for the component; helpful for visualization & debugging
    type: ErnieRanker    # pipelines Class name for the component
    params:
      model_name_or_path: rocketqa-nano-cross-encoder
      top_k: 3
  - name: TextFileConverter
    type: TextConverter
  - name: ImageFileConverter
    type: ImageToTextConverter
  - name: PDFFileConverter
    type: PDFToTextConverter
  - name: DocxFileConverter
    type: DocxToTextConverter
  - name: Preprocessor
    type: PreProcessor
    params:
      split_by: word
      split_length: 1000
  - name: FileTypeClassifier
    type: FileTypeClassifier

pipelines:
  - name: query    # a sample extractive-qa Pipeline
    type: Query
    nodes:
      - name: Retriever
        inputs: [Query]
      - name: Ranker
        inputs: [Retriever]
  - name: indexing
    type: Indexing
    nodes:
      - name: FileTypeClassifier
        inputs: [File]
      - name: TextFileConverter
        inputs: [FileTypeClassifier.output_1]
      - name: PDFFileConverter
        inputs: [FileTypeClassifier.output_2]
      - name: DocxFileConverter
        inputs: [FileTypeClassifier.output_4]
      - name: ImageFileConverter
        inputs: [FileTypeClassifier.output_6]
      - name: Preprocessor
        inputs: [PDFFileConverter, TextFileConverter, DocxFileConverter, ImageFileConverter]
      - name: Retriever
        inputs: [Preprocessor]
      - name: DocumentStore
        inputs: [Retriever]

[w5688414]

有具体的数据不？我们评估的是base比nano强，您可以再检查一下