Notebook

递归检索器 + 节点引用¶

本指南展示了如何使用递归检索来遍历节点关系，并根据“引用”获取节点。

节点引用是一个强大的概念。当您首次执行检索时，您可能希望检索引用而不是原始文本。您可以有多个引用指向同一个节点。

在本指南中，我们探讨了节点引用的一些不同用法：

块引用：不同块大小指向一个更大的块
元数据引用：摘要 + 生成的问题指向一个更大的块

In [ ]:

%pip install llama-index-llms-openai
%pip install llama-index-readers-file

In [ ]:

%load_ext autoreload
%autoreload 2
%env OPENAI_API_KEY=YOUR_OPENAI_KEY

如果您在colab上打开这个笔记本，您可能需要安装LlamaIndex 🦙。

In [ ]:

!pip install llama-index pypdf

加载数据 + 设置¶

在这一部分，我们将下载 Llama 2 论文并创建一个初始节点集（块大小为 1024）。

In [ ]:

!mkdir -p 'data/'
!wget --user-agent "Mozilla" "https://arxiv.org/pdf/2307.09288.pdf" -O "data/llama2.pdf"

Will not apply HSTS. The HSTS database must be a regular and non-world-writable file.
ERROR: could not open HSTS store at '/home/loganm/.wget-hsts'. HSTS will be disabled.
--2024-01-01 11:13:01--  https://arxiv.org/pdf/2307.09288.pdf
Resolving arxiv.org (arxiv.org)... 151.101.3.42, 151.101.131.42, 151.101.67.42, ...
Connecting to arxiv.org (arxiv.org)|151.101.3.42|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 13661300 (13M) [application/pdf]
Saving to: ‘data/llama2.pdf’

data/llama2.pdf     100%[===================>]  13.03M  27.3MB/s    in 0.5s    

2024-01-01 11:13:02 (27.3 MB/s) - ‘data/llama2.pdf’ saved [13661300/13661300]

In [ ]:

from pathlib import Path
from llama_index.readers.file import PDFReader
from llama_index.core.response.notebook_utils import display_source_node
from llama_index.core.retrievers import RecursiveRetriever
from llama_index.core.query_engine import RetrieverQueryEngine
from llama_index.core import VectorStoreIndex
from llama_index.llms.openai import OpenAI
import json

In [ ]:

loader = PDFReader()
docs0 = loader.load_data(file=Path("./data/llama2.pdf"))

In [ ]:

from llama_index.core import Document

doc_text = "\n\n".join([d.get_content() for d in docs0])
docs = [Document(text=doc_text)]

In [ ]:

from llama_index.core.node_parser import SentenceSplitter
from llama_index.core.schema import IndexNode

In [ ]:

node_parser = SentenceSplitter(chunk_size=1024)

In [ ]:

base_nodes = node_parser.get_nodes_from_documents(docs)# 将节点id设置为一个常量for idx, node in enumerate(base_nodes):    node.id_ = f"node-{idx}"

In [ ]:

from llama_index.core.embeddings import resolve_embed_model

embed_model = resolve_embed_model("local:BAAI/bge-small-en")
llm = OpenAI(model="gpt-3.5-turbo")

基准检索器¶

定义一个基准检索器，它通过嵌入相似度简单地获取前k个原始文本节点。

In [ ]:

base_index = VectorStoreIndex(base_nodes, embed_model=embed_model)
base_retriever = base_index.as_retriever(similarity_top_k=2)

In [ ]:

retrievals = base_retriever.retrieve(
    "Can you tell me about the key concepts for safety finetuning"
)

In [ ]:

for n in retrievals:
    display_source_node(n, source_length=1500)

Node ID: node-26
Similarity: 0.8581930837671874
Text: AsLLMsareintegratedanddeployed,welookforwardto continuing research that will amplify their potential for positive impact on these important social issues. 4.2 Safety Fine-Tuning In this section, we describe our approach to safety fine-tuning, including safety categories, annotation guidelines,andthetechniquesweusetomitigatesafetyrisks. Weemployaprocesssimilartothegeneral fine-tuning methods as described in Section 3, with some notable differences related to safety concerns. Specifically, we use the following techniques in safety fine-tuning: 1.Supervised Safety Fine-Tuning : We initialize by gathering adversarial prompts and safe demonstra- tions that are then included in the general supervised fine-tuning process (Section 3.1). This teaches themodeltoalignwithoursafetyguidelinesevenbeforeRLHF,andthuslaysthefoundationfor high-quality human preference data annotation. 2.Safety RLHF : Subsequently, we integrate safety in the general RLHF pipeline described in Sec- tion 3.2.2. This includes training a safety-specific reward model and gathering more challenging adversarial prompts for rejection sampling style fine-tuning and PPO optimization. 3.SafetyContextDistillation : Finally,werefineourRLHFpipelinewithcontextdistillation(Askell etal.,2021b). Thisinvolvesgeneratingsafermodelresponsesbyprefixingapromptwithasafety preprompt, e.g., “You are a safe and responsible assistant,” and then fine-tuning the model on the safer responses without the preprompt, which essentially distill...

Node ID: node-33
Similarity: 0.8442806881920294
Text: Examples of evaluation prompts and responses can be found in Appendix A.4.6. We then asked raters to judge models for safety violations on a five-point Likert scale with the following definitions: •5- No safety violations and very helpful •4- No safety violations and only minor non-safety issues •3- No safety violations but not helpful or other major non-safety issues •2- Mild or moderate safety violations 29

(a)Overall violation percentage. (b)Overall safety and helpfulness mean rating. Figure17: Overallsafetymeasures. Left: Llama 2-Chat haslowviolationpercentageoverallacrossmodel sizes.Right: Llama 2-Chat has high safety and helpfulness mean rating overall across model sizes. It is important to note that these results are subject to limitations of the prompt set, subjectivity of the review guidelines, and subjectivity of individual raters. •1- Severe safety violations We consider a rating of 1 or 2 as violation and use violation percentage as our main evaluation metric, with themeanratingasasupplement. Eachexampleisannotatedbythreeannotatorsandwetakethemajority votetodetermineiftheresponseisviolatingornot. WeusedGwet’sAC1/2statistictomeasureinter-rater reliability(IRR)asinthehelpfulnesshumanevaluation. TheIRRscoresrangefrom 0.70to0.95depending on the annotation batch, indicating a high degree of agreement among annotators on safety assessments. OnLlama 2-Chat annotations, the average IRR is 0.92according to Gwet’s AC2 measure. We see lower IRR scoresonbatcheswherethemo...

In [ ]:

query_engine_base = RetrieverQueryEngine.from_args(base_retriever, llm=llm)

In [ ]:

response = query_engine_base.query(
    "Can you tell me about the key concepts for safety finetuning"
)
print(str(response))

The key concepts for safety fine-tuning include supervised safety fine-tuning, safety RLHF (Reinforcement Learning from Human Feedback), and safety context distillation. In supervised safety fine-tuning, adversarial prompts and safe demonstrations are gathered and included in the general supervised fine-tuning process. This helps the model align with safety guidelines and lays the foundation for high-quality human preference data annotation. Safety RLHF involves integrating safety in the general RLHF pipeline, which includes training a safety-specific reward model and gathering more challenging adversarial prompts for rejection sampling style fine-tuning and PPO (Proximal Policy Optimization) optimization. Safety context distillation is the final step, where the RLHF pipeline is refined with context distillation. This involves generating safer model responses by prefixing a prompt with a safety preprompt and then fine-tuning the model on the safer responses without the preprompt.

块引用：较小的子块指向较大的父块¶

在这个用法示例中，我们将展示如何构建一个较小的子块指向较大的父块的图。

在查询时，我们检索较小的子块，但是我们跟随引用到较大的父块。这样可以让我们在合成时获得更多的上下文信息。

In [ ]:

sub_chunk_sizes = [128, 256, 512]sub_node_parsers = [    SentenceSplitter(chunk_size=c, chunk_overlap=20) for c in sub_chunk_sizes]all_nodes = []for base_node in base_nodes:    for n in sub_node_parsers:        sub_nodes = n.get_nodes_from_documents([base_node])        sub_inodes = [            IndexNode.from_text_node(sn, base_node.node_id) for sn in sub_nodes        ]        all_nodes.extend(sub_inodes)    # also add original node to node    original_node = IndexNode.from_text_node(base_node, base_node.node_id)    all_nodes.append(original_node)

In [ ]:

all_nodes_dict = {n.node_id: n for n in all_nodes}

In [ ]:

vector_index_chunk = VectorStoreIndex(all_nodes, embed_model=embed_model)

In [ ]:

vector_retriever_chunk = vector_index_chunk.as_retriever(similarity_top_k=2)

In [ ]:

retriever_chunk = RecursiveRetriever(
    "vector",
    retriever_dict={"vector": vector_retriever_chunk},
    node_dict=all_nodes_dict,
    verbose=True,
)

In [ ]:

nodes = retriever_chunk.retrieve(
    "Can you tell me about the key concepts for safety finetuning"
)
for node in nodes:
    display_source_node(node, source_length=2000)

Retrieving with query id None: Can you tell me about the key concepts for safety finetuning
Retrieved node with id, entering: node-26
Retrieving with query id node-26: Can you tell me about the key concepts for safety finetuning
Retrieved node with id, entering: node-1
Retrieving with query id node-1: Can you tell me about the key concepts for safety finetuning

Node ID: node-26
Similarity: 0.8809071991986446
Text: AsLLMsareintegratedanddeployed,welookforwardto continuing research that will amplify their potential for positive impact on these important social issues. 4.2 Safety Fine-Tuning In this section, we describe our approach to safety fine-tuning, including safety categories, annotation guidelines,andthetechniquesweusetomitigatesafetyrisks. Weemployaprocesssimilartothegeneral fine-tuning methods as described in Section 3, with some notable differences related to safety concerns. Specifically, we use the following techniques in safety fine-tuning: 1.Supervised Safety Fine-Tuning : We initialize by gathering adversarial prompts and safe demonstra- tions that are then included in the general supervised fine-tuning process (Section 3.1). This teaches themodeltoalignwithoursafetyguidelinesevenbeforeRLHF,andthuslaysthefoundationfor high-quality human preference data annotation. 2.Safety RLHF : Subsequently, we integrate safety in the general RLHF pipeline described in Sec- tion 3.2.2. This includes training a safety-specific reward model and gathering more challenging adversarial prompts for rejection sampling style fine-tuning and PPO optimization. 3.SafetyContextDistillation : Finally,werefineourRLHFpipelinewithcontextdistillation(Askell etal.,2021b). Thisinvolvesgeneratingsafermodelresponsesbyprefixingapromptwithasafety preprompt, e.g., “You are a safe and responsible assistant,” and then fine-tuning the model on the safer responses without the preprompt, which essentially distillsthe safety preprompt (context) into the model. Weuseatargetedapproachthatallowsoursafetyrewardmodeltochoosewhethertouse context distillation for each sample. 4.2.1 Safety Categories and Annotation Guidelines Based on limitations of LLMs known from prior work, we design instructions for our annotation team to createadversarialpromptsalongtwodimensions: a riskcategory ,orpotentialtopicaboutwhichtheLLM couldproduceunsafecontent;andan attackvector ,orquestionstyletocoverdifferentvarietiesofprompts ...

Node ID: node-1
Similarity: 0.8744334039911964
Text: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2 Reinforcement Learning with Human Feedback (RLHF) . . . . . . . . . . . . . . . . . . . . . 9 3.3 System Message for Multi-Turn Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.4 RLHF Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4 Safety 20 4.1 Safety in Pretraining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 4.2 Safety Fine-Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.3 Red Teaming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.4 Safety Evaluation of Llama 2-Chat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 5 Discussion 32 5.1 Learnings and Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 5.2 Limitations and Ethical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 5.3 Responsible Release Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 6 Related Work 35 7 Conclusion 36 A Appendix 46 A.1 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

In [ ]:

query_engine_chunk = RetrieverQueryEngine.from_args(retriever_chunk, llm=llm)

In [ ]:

response = query_engine_chunk.query(
    "Can you tell me about the key concepts for safety finetuning"
)
print(str(response))

Retrieving with query id None: Can you tell me about the key concepts for safety finetuning
Retrieved node with id, entering: node-26
Retrieving with query id node-26: Can you tell me about the key concepts for safety finetuning
Retrieved node with id, entering: node-1
Retrieving with query id node-1: Can you tell me about the key concepts for safety finetuning
The key concepts for safety fine-tuning include supervised safety fine-tuning, safety RLHF (Reinforcement Learning with Human Feedback), and safety context distillation. Supervised safety fine-tuning involves gathering adversarial prompts and safe demonstrations to teach the model to align with safety guidelines. Safety RLHF integrates safety into the general RLHF pipeline by training a safety-specific reward model and gathering challenging adversarial prompts for rejection sampling style fine-tuning and PPO optimization. Safety context distillation involves generating safer model responses by prefixing a prompt with a safety preprompt and fine-tuning the model on the safer responses without the preprompt. These techniques aim to mitigate safety risks and improve the model's ability to provide safe and responsible responses.

元数据引用：对更大块内容的摘要和生成的问题¶

在这个用法示例中，我们展示了如何定义引用源节点的附加上下文。

这个附加上下文包括摘要以及生成的问题。

在查询时，我们检索较小的块，但是我们会跟随引用到更大的块。这样可以为综合提供更多的上下文。

In [ ]:

import nest_asyncio

nest_asyncio.apply()

In [ ]:

from llama_index.core.node_parser import SentenceSplitter
from llama_index.core.schema import IndexNode
from llama_index.core.extractors import (
    SummaryExtractor,
    QuestionsAnsweredExtractor,
)

In [ ]:

extractors = [
    SummaryExtractor(summaries=["self"], show_progress=True),
    QuestionsAnsweredExtractor(questions=5, show_progress=True),
]

In [ ]:

# 在基本节点上运行元数据提取器，返回字典node_to_metadata = {}for extractor in extractors:    metadata_dicts = extractor.extract(base_nodes)    for node, metadata in zip(base_nodes, metadata_dicts):        if node.node_id not in node_to_metadata:            node_to_metadata[node.node_id] = metadata        else:            node_to_metadata[node.node_id].update(metadata)

100%|██████████| 93/93 [01:13<00:00,  1.27it/s]
100%|██████████| 93/93 [00:49<00:00,  1.88it/s]

In [ ]:

# 缓存元数据字典def save_metadata_dicts(path, data):    with open(path, "w") as fp:        json.dump(data, fp)def load_metadata_dicts(path):    with open(path, "r") as fp:        data = json.load(fp)    return data

In [ ]:

save_metadata_dicts("data/llama2_metadata_dicts.json", node_to_metadata)

In [ ]:

metadata_dicts = load_metadata_dicts("data/llama2_metadata_dicts.json")

In [ ]:

# 所有节点都由源节点和元数据组成import copyall_nodes = copy.deepcopy(base_nodes)for node_id, metadata in node_to_metadata.items():    for val in metadata.values():        all_nodes.append(IndexNode(text=val, index_id=node_id))

In [ ]:

all_nodes_dict = {n.node_id: n for n in all_nodes}

In [ ]:

## 将索引加载到向量索引中from llama_index.core import VectorStoreIndexfrom llama_index.llms.openai import OpenAIllm = OpenAI(model="gpt-3.5-turbo")vector_index_metadata = VectorStoreIndex(all_nodes)

In [ ]:

vector_retriever_metadata = vector_index_metadata.as_retriever(
    similarity_top_k=2
)

In [ ]:

retriever_metadata = RecursiveRetriever(
    "vector",
    retriever_dict={"vector": vector_retriever_metadata},
    node_dict=all_nodes_dict,
    verbose=False,
)

In [ ]:

nodes = retriever_metadata.retrieve(
    "Can you tell me about the key concepts for safety finetuning"
)
for node in nodes:
    display_source_node(node, source_length=2000)

Node ID: node-26
Similarity: 0.8727061238826861
Text: AsLLMsareintegratedanddeployed,welookforwardto continuing research that will amplify their potential for positive impact on these important social issues. 4.2 Safety Fine-Tuning In this section, we describe our approach to safety fine-tuning, including safety categories, annotation guidelines,andthetechniquesweusetomitigatesafetyrisks. Weemployaprocesssimilartothegeneral fine-tuning methods as described in Section 3, with some notable differences related to safety concerns. Specifically, we use the following techniques in safety fine-tuning: 1.Supervised Safety Fine-Tuning : We initialize by gathering adversarial prompts and safe demonstra- tions that are then included in the general supervised fine-tuning process (Section 3.1). This teaches themodeltoalignwithoursafetyguidelinesevenbeforeRLHF,andthuslaysthefoundationfor high-quality human preference data annotation. 2.Safety RLHF : Subsequently, we integrate safety in the general RLHF pipeline described in Sec- tion 3.2.2. This includes training a safety-specific reward model and gathering more challenging adversarial prompts for rejection sampling style fine-tuning and PPO optimization. 3.SafetyContextDistillation : Finally,werefineourRLHFpipelinewithcontextdistillation(Askell etal.,2021b). Thisinvolvesgeneratingsafermodelresponsesbyprefixingapromptwithasafety preprompt, e.g., “You are a safe and responsible assistant,” and then fine-tuning the model on the safer responses without the preprompt, which essentially distillsthe safety preprompt (context) into the model. Weuseatargetedapproachthatallowsoursafetyrewardmodeltochoosewhethertouse context distillation for each sample. 4.2.1 Safety Categories and Annotation Guidelines Based on limitations of LLMs known from prior work, we design instructions for our annotation team to createadversarialpromptsalongtwodimensions: a riskcategory ,orpotentialtopicaboutwhichtheLLM couldproduceunsafecontent;andan attackvector ,orquestionstyletocoverdifferentvarietiesofprompts ...

Node ID: node-26
Similarity: 0.8586079224453517
Text: AsLLMsareintegratedanddeployed,welookforwardto continuing research that will amplify their potential for positive impact on these important social issues. 4.2 Safety Fine-Tuning In this section, we describe our approach to safety fine-tuning, including safety categories, annotation guidelines,andthetechniquesweusetomitigatesafetyrisks. Weemployaprocesssimilartothegeneral fine-tuning methods as described in Section 3, with some notable differences related to safety concerns. Specifically, we use the following techniques in safety fine-tuning: 1.Supervised Safety Fine-Tuning : We initialize by gathering adversarial prompts and safe demonstra- tions that are then included in the general supervised fine-tuning process (Section 3.1). This teaches themodeltoalignwithoursafetyguidelinesevenbeforeRLHF,andthuslaysthefoundationfor high-quality human preference data annotation. 2.Safety RLHF : Subsequently, we integrate safety in the general RLHF pipeline described in Sec- tion 3.2.2. This includes training a safety-specific reward model and gathering more challenging adversarial prompts for rejection sampling style fine-tuning and PPO optimization. 3.SafetyContextDistillation : Finally,werefineourRLHFpipelinewithcontextdistillation(Askell etal.,2021b). Thisinvolvesgeneratingsafermodelresponsesbyprefixingapromptwithasafety preprompt, e.g., “You are a safe and responsible assistant,” and then fine-tuning the model on the safer responses without the preprompt, which essentially distillsthe safety preprompt (context) into the model. Weuseatargetedapproachthatallowsoursafetyrewardmodeltochoosewhethertouse context distillation for each sample. 4.2.1 Safety Categories and Annotation Guidelines Based on limitations of LLMs known from prior work, we design instructions for our annotation team to createadversarialpromptsalongtwodimensions: a riskcategory ,orpotentialtopicaboutwhichtheLLM couldproduceunsafecontent;andan attackvector ,orquestionstyletocoverdifferentvarietiesofprompts ...

In [ ]:

query_engine_metadata = RetrieverQueryEngine.from_args(
    retriever_metadata, llm=llm
)

In [ ]:

response = query_engine_metadata.query(
    "Can you tell me about the key concepts for safety finetuning"
)
print(str(response))

The key concepts for safety fine-tuning include supervised safety fine-tuning, safety RLHF (Reinforcement Learning from Human Feedback), and safety context distillation. Supervised safety fine-tuning involves gathering adversarial prompts and safe demonstrations to train the model to align with safety guidelines. Safety RLHF integrates safety into the RLHF pipeline by training a safety-specific reward model and gathering challenging adversarial prompts for fine-tuning and optimization. Safety context distillation involves generating safer model responses by prefixing a prompt with a safety preprompt and fine-tuning the model on the safer responses without the preprompt. These concepts are used to mitigate safety risks and improve the model's ability to produce safe and helpful responses.

评估¶

我们评估我们的递归检索 + 节点引用方法的效果。我们评估块引用和元数据引用两种情况。我们使用嵌入相似度查找来检索引用节点。

我们将这两种方法与直接获取原始节点的基准检索器进行比较。

在指标方面，我们使用命中率和MRR进行评估。

数据集生成¶

我们首先从文本块集合中生成一个问题数据集。

In [ ]:

from llama_index.core.evaluation import (
    generate_question_context_pairs,
    EmbeddingQAFinetuneDataset,
)
from llama_index.llms.openai import OpenAI

import nest_asyncio

nest_asyncio.apply()

In [ ]:

eval_dataset = generate_question_context_pairs(
    base_nodes, OpenAI(model="gpt-3.5-turbo")
)

100%|██████████| 93/93 [02:08<00:00,  1.38s/it]

In [ ]:

eval_dataset.save_json("data/llama2_eval_dataset.json")

In [ ]:

# 可选eval_dataset = EmbeddingQAFinetuneDataset.from_json(    "data/llama2_eval_dataset.json")

比较结果¶

我们对每个检索器进行评估，以衡量命中率和MRR。

我们发现，具有节点引用（无论是块还是元数据）的检索器往往比检索原始块表现更好。

In [ ]:

import pandas as pdfrom llama_index.core.evaluation import (    RetrieverEvaluator,    get_retrieval_results_df,)# 将向量检索相似度的 top k 设置为较高值top_k = 10def display_results(names, results_arr):    """显示评估结果。"""    hit_rates = []    mrrs = []    for name, eval_results in zip(names, results_arr):        metric_dicts = []        for eval_result in eval_results:            metric_dict = eval_result.metric_vals_dict            metric_dicts.append(metric_dict)        results_df = pd.DataFrame(metric_dicts)        hit_rate = results_df["hit_rate"].mean()        mrr = results_df["mrr"].mean()        hit_rates.append(hit_rate)        mrrs.append(mrr)    final_df = pd.DataFrame(        {"retrievers": names, "hit_rate": hit_rates, "mrr": mrrs}    )    display(final_df)

In [ ]:

vector_retriever_chunk = vector_index_chunk.as_retriever(    similarity_top_k=top_k)retriever_chunk = RecursiveRetriever(    "vector",    retriever_dict={"vector": vector_retriever_chunk},    node_dict=all_nodes_dict,    verbose=True,)retriever_evaluator = RetrieverEvaluator.from_metric_names(    ["mrr", "hit_rate"], retriever=retriever_chunk)# 在整个数据集上尝试results_chunk = await retriever_evaluator.aevaluate_dataset(    eval_dataset, show_progress=True)

In [ ]:

# 将向量检索器元数据设置为与top_k相似度的向量索引元数据vector_retriever_metadata = vector_index_metadata.as_retriever(    similarity_top_k=top_k)# 创建递归检索器元数据retriever_metadata = RecursiveRetriever(    "vector",    retriever_dict={"vector": vector_retriever_metadata},    node_dict=all_nodes_dict,    verbose=True,)# 从指标名称创建检索器评估器retriever_evaluator = RetrieverEvaluator.from_metric_names(    ["mrr", "hit_rate"], retriever=retriever_metadata)# 尝试在整个数据集上进行评估results_metadata = await retriever_evaluator.evaluate_dataset(    eval_dataset, show_progress=True)

In [ ]:

# 将基础检索器转换为检索器base_retriever = base_index.as_retriever(similarity_top_k=top_k)retriever_evaluator = RetrieverEvaluator.from_metric_names(    ["mrr", "hit_rate"], retriever=base_retriever)# 在整个数据集上尝试results_base = await retriever_evaluator.aevaluate_dataset(    eval_dataset, show_progress=True)

100%|██████████| 194/194 [00:09<00:00, 19.86it/s]

In [ ]:

full_results_df = get_retrieval_results_df(
    [
        "Base Retriever",
        "Retriever (Chunk References)",
        "Retriever (Metadata References)",
    ],
    [results_base, results_chunk, results_metadata],
)
display(full_results_df)

	retrievers	hit_rate	mrr
0	Base Retriever	0.778351	0.563103
1	Retriever (Chunk References)	0.896907	0.691114
2	Retriever (Metadata References)	0.891753	0.718440