这是一份全面的 RAG 指南,详细阐述了采用 RAG 的动机,以及如何超越基础或初级 RAG 构建的技术和策略。(高清版本链接)
新的一年伊始,你可能正考虑进入 RAG 领域,尝试构建你的首个 RAG 系统。或者,你已经构建了基础 RAG 系统,现在希望进一步提升,以便更好地处理用户的查询和数据结构。
无论你处于哪种情况,如何着手可能都是一个挑战!希望这篇博客文章能为你指明下一步的方向,并为你在构建高级 RAG 系统时提供一个思维模型,帮助你做出决策。
上文提到的 RAG 指南,很大程度上是受到了最近的一篇 RAG 综述论文的启发(“Retrieval-Augmented Generation for Large Language Models: A Survey”Gao, Yunfan 等人,2023)。
基础 RAG
今天的主流 RAG 涉及从外部知识库检索文档,并将这些文档及用户的查询传递给大语言模型(LLM),以生成响应。换言之,RAG 包含了一个检索组件、一个外部知识库和一个生成组件。
LlamaIndex 基础 RAG 指南:
from llama_index import SimpleDirectoryReader, VectorStoreIndex
# load data
documents = SimpleDirectoryReader(input_dir="...").load_data()
# build VectorStoreIndex that takes care of chunking documents
# and encoding chunks to embeddings for future retrieval
index = VectorStoreIndex.from_documents(documents=documents)
# The QueryEngine class is equipped with the generator
# and facilitates the retrieval and generation steps
query_engine = index.as_query_engine()
# Use your Default RAG
response = query_engine.query("A user's query")
RAG 成功的要求
为了使 RAG 系统成功(即能够为用户问题提供有用且相关的答案),主要有两个高层次的要求:
- 检索组件必须能够找到与用户查询最相关的文档。
- 生成组件必须能够有效利用检索到的文档,充分回答用户的查询。
高级 RAG
在明确了成功的要求后,我们可以说,构建高级 RAG 实际上是关于运用更复杂的技术和策略(应用于检索或生成组件),以确保这些要求得以满足。此外,我们可以将复杂的技术归类为:要么独立地解决两个高层次成功要求中的一个,要么同时解决这两个要求。
如何找到与用户查询最相关的文档:高级检索技术探索
接下来,我们将简要介绍几种复杂但有效的技术,以帮助实现有效检索的首要目标。
- 优化文档分块大小 (Chunk-Size Optimization): 由于大语言模型 (LLM) 的上下文长度限制,我们在构建外部知识库时必须对文档进行分块。块大小不当会影响生成响应的准确性,因此这一步骤至关重要。
LlamaIndex 文档块大小优化方法 (LlamaIndex Chunk Size Optimization Recipe) (教程 (notebook guide)):
from llama_index import ServiceContext
from llama_index.param_tuner.base import ParamTuner, RunResult
from llama_index.evaluation import SemanticSimilarityEvaluator, BatchEvalRunner
### Recipe
### Perform hyperparameter tuning as in traditional ML via grid-search
### 1. Define an objective function that ranks different parameter combos
### 2. Build ParamTuner object
### 3. Execute hyperparameter tuning with ParamTuner.tune()
# 1. Define objective function
defobjective_function(params_dict):
chunk_size = params_dict["chunk_size"]
docs = params_dict["docs"]
top_k = params_dict["top_k"]
eval_qs = params_dict["eval_qs"]
ref_response_strs = params_dict["ref_response_strs"]
# build RAG pipeline
index = _build_index(chunk_size, docs)# helper function not shown here
query_engine = index.as_query_engine(similarity_top_k=top_k)
# perform inference with RAG pipeline on a provided questions `eval_qs`
pred_response_objs = get_responses(
eval_qs, query_engine, show_progress=True
)
# perform evaluations of predictions by comparing them to reference
# responses `ref_response_strs`
evaluator = SemanticSimilarityEvaluator(...)
eval_batch_runner = BatchEvalRunner(
{"semantic_similarity": evaluator}, workers=2, show_progress=True
)
eval_results = eval_batch_runner.evaluate_responses(
eval_qs, responses=pred_response_objs, reference=ref_response_strs
)
# get semantic similarity metric
mean_score = np.array(
[r.score for r in eval_results["semantic_similarity"]]
).mean()
return RunResult(score=mean_score, params=params_dict)
# 2. Build ParamTuner object
param_dict ={"chunk_size":[256,512,1024]}# params/values to search over
fixed_param_dict ={# fixed hyperparams
"top_k":2,
"docs": docs,
"eval_qs": eval_qs[:10],
"ref_response_strs": ref_response_strs[:10],
}
param_tuner = ParamTuner(
param_fn=objective_function,
param_dict=param_dict,
fixed_param_dict=fixed_param_dict,
show_progress=True,
)
# 3. Execute hyperparameter search
results = param_tuner.tune()
best_result = results.best_run_result
best_chunk_size = results.best_run_result.params["chunk_size"]
2. 构建结构化外部知识 (Structured External Knowledge): 面对复杂场景时,比起普通的向量索引,我们可能需要一个更有结构性的外部知识库。这样的设计可以在处理分散的知识源时,实现更精准的递归检索或路由检索。
LlamaIndex 结构化检索方法 (LlamaIndex Recursive Retrieval Recipe) (教程 (notebook guide)):
from llama_index import SimpleDirectoryReader, VectorStoreIndex
from llama_index.node_parser import SentenceSplitter
from llama_index.schema import IndexNode
### Recipe
### Build a recursive retriever that retrieves using small chunks
### but passes associated larger chunks to the generation stage
# load data
documents = SimpleDirectoryReader(
input_file="some_data_path/llama2.pdf"
).load_data()
# build parent chunks via NodeParser
node_parser = SentenceSplitter(chunk_size=1024)
base_nodes = node_parser.get_nodes_from_documents(documents)
# define smaller child chunks
sub_chunk_sizes =[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)
# define a VectorStoreIndex with all of the nodes
vector_index_chunk = VectorStoreIndex(
all_nodes, service_context=service_context
)
vector_retriever_chunk = vector_index_chunk.as_retriever(similarity_top_k=2)
# build RecursiveRetriever
all_nodes_dict ={n.node_id: n for n in all_nodes}
retriever_chunk = RecursiveRetriever(
"vector",
retriever_dict={"vector": vector_retriever_chunk},
node_dict=all_nodes_dict,
verbose=True,
)
# build RetrieverQueryEngine using recursive_retriever
query_engine_chunk = RetrieverQueryEngine.from_args(
retriever_chunk, service_context=service_context
)
# perform inference with advanced RAG (i.e. query engine)
response = query_engine_chunk.query(
"Can you tell me about the key concepts for safety finetuning"
)
其他推荐资源
为了在复杂的检索情况下实现高准确度,我们准备了一系列高级技术的应用指南。以下是其中一些精选教程的链接:
- 利用知识图谱构建外部知识库 (Building External Knowledge using Knowledge Graphs)
- 结合自动检索器实现混合式检索 (Performing Mixed Retrieval with Auto Retrievers)
- 创建融合检索器 (Building Fusion Retrievers)
- 优化检索中使用的嵌入模型 (Fine-tuning Embedding Models used in Retrieval)
- 改进查询嵌入的方法 (HyDE) (Transforming Query Embeddings (HyDE))
高级生成技术必须高效利用检索到的文档
本节内容与前一节相似,我们将展示一些高级技术的例子。这些技术的核心在于确保检索到的文档与生成器使用的大语言模型 (LLM) 高度匹配。
- 信息压缩: 大语言模型在处理信息时受到上下文长度的限制。此外,如果检索到的文档含有过多无关信息(即“噪音”),会导致生成的回应质量下降。
LlamaIndex 信息压缩方法(请参阅笔记本指南):
from llama_index import SimpleDirectoryReader, VectorStoreIndex
from llama_index.query_engine import RetrieverQueryEngine
from llama_index.postprocessor import LongLLMLinguaPostprocessor
### Recipe
### Define a Postprocessor object, here LongLLMLinguaPostprocessor
### Build QueryEngine that uses this Postprocessor on retrieved docs
# Define Postprocessor
node_postprocessor = LongLLMLinguaPostprocessor(
instruction_str="Given the context, please answer the final question",
target_token=300,
rank_method="longllmlingua",
additional_compress_kwargs={
"condition_compare":True,
"condition_in_question":"after",
"context_budget":"+100",
"reorder_context":"sort",# enable document reorder
},
)
# Define VectorStoreIndex
documents = SimpleDirectoryReader(input_dir="...").load_data()
index = VectorStoreIndex.from_documents(documents)
# Define QueryEngine
retriever = index.as_retriever(similarity_top_k=2)
retriever_query_engine = RetrieverQueryEngine.from_args(
retriever, node_postprocessors=[node_postprocessor]
)
# Used your advanced RAG
response = retriever_query_engine.query("A user query")
- 结果重新排序: 大语言模型存在一种被称为“中途迷失”现象,即模型倾向于只关注提示语两端的极端内容。因此,在将文档提交给生成组件前,对其重新排序可以提高生成内容的质量。
LlamaIndex 结果重排序改进生成方法(请参阅笔记本指南):
import os
from llama_index import SimpleDirectoryReader, VectorStoreIndex
from llama_index.postprocessor.cohere_rerank import CohereRerank
from llama_index.postprocessor import LongLLMLinguaPostprocessor
### Recipe
### Define a Postprocessor object, here CohereRerank
### Build QueryEngine that uses this Postprocessor on retrieved docs
# Build CohereRerank post retrieval processor
api_key = os.environ["COHERE_API_KEY"]
cohere_rerank = CohereRerank(api_key=api_key, top_n=2)
# Build QueryEngine (RAG) using the post processor
documents = SimpleDirectoryReader("./data/paul_graham/").load_data()
index = VectorStoreIndex.from_documents(documents=documents)
query_engine = index.as_query_engine(
similarity_top_k=10,
node_postprocessors=[cohere_rerank],
)
# Use your advanced RAG
response = query_engine.query(
"What did Sam Altman do in this essay?"
)
高级技术用于同时提升检索和生成效果
在这个小节中,我们探讨了一些同时考虑检索与生成相结合的复杂技术,以期实现更有效的检索和更准确的生成回应。
- 生成器增强的检索: 这些技术利用大语言模型固有的推理能力,在检索前先对用户的查询进行精细化处理,从而更准确地确定所需的信息,以提供有效的回应。
LlamaIndex 生成器增强检索方法(请参阅笔记本指南):
from llama_index.llms import OpenAI
from llama_index.query_engine import FLAREInstructQueryEngine
from llama_index import(
VectorStoreIndex,
SimpleDirectoryReader,
ServiceContext,
)
### Recipe
### Build a FLAREInstructQueryEngine which has the generator LLM play
### a more active role in retrieval by prompting it to elicit retrieval
### instructions on what it needs to answer the user query.
# Build FLAREInstructQueryEngine
documents = SimpleDirectoryReader("./data/paul_graham").load_data()
index = VectorStoreIndex.from_documents(documents)
index_query_engine = index.as_query_engine(similarity_top_k=2)
service_context = ServiceContext.from_defaults(llm=OpenAI(model="gpt-4"))
flare_query_engine = FLAREInstructQueryEngine(
query_engine=index_query_engine,
service_context=service_context,
max_iterations=7,
verbose=True,
)
# Use your advanced RAG
response = flare_query_engine.query(
"Can you tell me about the author's trajectory in the startup world?"
)
- 迭代式检索与生成器相结合的 RAG: 在一些复杂的情况下,可能需要多步骤的推理来提供与用户查询相关且有用的答案。
LlamaIndex 迭代式检索与生成器结合方法(请参阅笔记本指南):
from llama_index.query_engine import RetryQueryEngine
from llama_index.evaluation import RelevancyEvaluator
### Recipe
### Build a RetryQueryEngine which performs retrieval-generation cycles
### until it either achieves a passing evaluation or a max number of
### cycles has been reached
# Build RetryQueryEngine
documents = SimpleDirectoryReader("./data/paul_graham").load_data()
index = VectorStoreIndex.from_documents(documents)
base_query_engine = index.as_query_engine()
query_response_evaluator = RelevancyEvaluator()# evaluator to critique
# retrieval-generation cycles
retry_query_engine = RetryQueryEngine(
base_query_engine, query_response_evaluator
)
# Use your advanced rag
retry_response = retry_query_engine.query("A user query")
RAG 测量方面的考量
对于 RAG 系统的评估自然是极其重要的。在 Gao, Yunfan 等人的调查论文中,他们提到了在 RAG 快速参考指南右上角所展示的 7 个关键测量方面。llama-index 库包括了多种评估工具和与 RAGAs 的整合功能,这些工具旨在帮助开发者从这些测量方面来评估他们的 RAG 系统是否满足预设的成功标准。下面,我们简要介绍了一些评估指南中的精选内容。
- 答案相关性和上下文相关性
- 内容的忠实性
- 信息检索效果的评估
- 使用批量评估工具 BatchEvalRunner 进行的评估
现在你已经准备好掌握高级 RAG 技术
希望通过阅读这篇博客文章,你能对使用这些先进技术构建高级 RAG 系统感到更加得心应手和充满信心!