|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/practice-exams-databricks-certified-generative-ai-engineer/
课程评论：没有评论
**Databricks Certified Generative AI Engineer Associate 模拟考试课程总结** 本课程提供两套模拟考试，旨在帮助您为“Databricks Certified Generative AI Engineer Associate”认证考试做好准备。每套模拟考试包含45道选择题（单选或多选），考试时间为90分钟。 **考试重点涵盖以下六大领域：** 1. **设计应用程序 (14%)**： * 设计有效的提示词以获得特定格式的响应。 * 根据业务需求选择合适的模型任务。 * 选择链式组件以满足模型的输入输出需求。 * 将业务用例目标转化为 AI 管道的输入输出描述。 * 定义和排序多阶段推理所需的工具。 2. **数据准备 (14%)**： * 根据文档结构和模型约束应用分块（chunking）策略。 * 过滤 RAG 应用中会降低质量的无关内容。 * 选择合适的 Python 包以提取和格式化文档内容。 * 定义在 Unity Catalog 的 Delta Lake 表中写入分块文本的操作顺序。 * 识别 RAG 应用所需知识集的来源文档。 * 识别与模型任务相符的提示/响应对。 * 使用工具和指标评估检索性能。 3. **应用程序开发 (30%)**： * 创建数据检索所需工具。 * 在生成式 AI 应用中选择 Langchain/类似工具。 * 识别提示格式如何影响模型输出。 * 定性评估响应，识别质量和安全问题。 * 基于模型和检索评估选择分块策略。 * 根据关键字段、术语和意图，使用额外上下文增强提示。 * 创建调整 LLM 响应的提示。 * 实现 LLM 护栏以防止负面结果。 * 编写最小化“幻觉”或泄露私有数据的元（meta）提示。 * 构建暴露可用功能的代理（agent）提示模板。 * 根据应用属性选择最佳 LLM。 * 根据源文档、预期查询和优化策略选择嵌入模型上下文长度。 * 根据模型元数据/模型卡片，从模型中心/市场为任务选择模型。 * 根据实验生成的常用指标，为给定任务选择最佳模型。 4. **组装和部署应用程序 (22%)**： * 使用带预处理和后处理的 pyfunc 模型编写链式（chain）。 * 控制模型服务终端对资源的访问。 * 根据要求编写简单的链式。 * 使用 Langchain 编写简单的链式。 * 创建 RAG 应用所需的基本元素：模型风味（flavor）、嵌入模型、检索器、依赖项、输入示例、模型签名。 * 使用 MLflow 将模型注册到 Unity Catalog。 * 部署基本 RAG 应用终端的步骤。 * 创建和查询向量搜索索引。 * 识别托管利用 Foundation Model API 的 LLM 应用的方法。 * 识别托管 RAG 应用所需功能的资源。 5. **治理 (8%)**： * 使用掩码（masking）技术作为护栏以达到性能目标。 * 选择护栏技术以防止恶意用户输入。 * 推荐用于 RAG 数据源的替代文本缓解方法。 * 利用数据源的法律/许可要求避免法律风险。 6. **评估和监控 (12%)**： * 根据定量评估指标选择 LLM（大小和架构）。 * 选择关键指标以监控特定的 LLM 部署场景。 * 使用 MLflow 评估 RAG 应用中的模型性能。 * 使用推理日志评估已部署 RAG 应用的性能。 * 使用 Databricks 功能控制 RAG 应用的 LLM 成本。 完成这些模拟考试，您将更有信心和能力在第一次尝试中通过 Databricks Certified Generative AI Engineer Associate 考试。祝您考试顺利！
Practice Tests: Databricks Certified Generative AI Engineer AssociateDescription: If you looking for practice tests for Databricks Certified Generative AI Engineer Associate exam, you have come to the right place! Two practice tests with detailed explanations are available to prepare you before appearing for the actual exam.About the Exam:1. Number of items: 45 multiple-choice or multiple-selection questions2. Time Limit: 90 minutes3. Registration fee: $2004. Delivery method: Online Proctored5. Validity: 2 years.8. Recertication: Recertication is required every two years to maintain your certified status.The practice tests cover the following exam topics with explanations:Section 1: Design ApplicationsDesign a prompt that elicits a specifically formatted responseSelect model tasks to accomplish a given business requirementSelect chain components for a desired model input and outputTranslate business use case goals into a description of the desired inputs and outputs for the AI pipelineDene and order tools that gather knowledge or take actions for multi-stage reasoningSection 2: Data PreparationApply a chunking strategy for a given document structure and model constraintsFilter extraneous content in source documents that degrades quality of a RAG applicationChoose the appropriate Python package to extract document content from provided source data and format.Dene operations and sequence to write given chunked text into Delta Lake tables in Unity CatalogIdentify needed source documents that provide necessary knowledge and quality for a given RAG applicationIdentify prompt/response pairs that align with a given model taskUse tools and metrics to evaluate retrieval performanceSection 3: Application DevelopmentCreate tools needed to extract data for a given data retrieval needSelect Langchain/similar tools for use in a Generative AI application.Identify how prompt formats can change model outputs and resultsQualitatively assess responses to identify common issues such as quality and safetySelect chunking strategy based on model & retrieval evaluationAugment a prompt with additional context from a user's input based on key elds, terms, and intentsCreate a prompt that adjusts an LLM's response from a baseline to a desired outputImplement LLM guardrails to prevent negative outcomesWrite metaprompts that minimize hallucinations or leaking private dataBuild agent prompt templates exposing available functionsSelect the best LLM based on the attributes of the application to be developedSelect a embedding model context length based on source documents, expected queries, and optimization strategySelect a model for from a model hub or marketplace for a task based on model metadata/model cardsSelect the best model for a given task based on common metrics generated in experimentsSection 4: Assembling and Deploying ApplicationsCode a chain using a pyfunc model with pre- and post-processingControl access to resources from model serving endpointsCode a simple chain according to requirementsCode a simple chain using langchainChoose the basic elements needed to create a RAG application: model avor, embedding model, retriever, dependencies, input examples, model signatureRegister the model to Unity Catalog using MLowSequence the steps needed to deploy an endpoint for a basic RAG applicationCreate and query a Vector Search indexIdentify how to serve an LLM application that leverages Foundation Model APIsIdentify resources needed to serve features for a RAG applicationSection 5: GovernanceUse masking techniques as guard rails to meet a performance objectiveSelect guardrail techniques to protect against malicious user inputs to a Gen AI application ● Recommend an alternative for problematic text mitigation in a data source feeding a RAG applicationUse legal/licensing requirements for data sources to avoid legal riskSection 6: Evaluation and MonitoringSelect an LLM choice (size and architecture) based on a set of quantitative evaluation metricsSelect key metrics to monitor for a specic LLM deployment scenarioEvaluate model performance in a RAG application using MLowUse inference logging to assess deployed RAG application performanceUse Databricks features to control LLM costs for RAG applicationsQuestions: There are 45 multiple-choice questions on each practice exam. The questions will be distributed topic wise in the following way:1. Design Applications - 14%2. Data Preparation - 14%3. Application Development - 30%4. Assembling and Deploying Apps - 22%5. Governance - 8%6. Evaluation and Monitoring - 12%By completing these practice tests, you will gain the confidence and knowledge needed to pass the Databricks Certified Generative AI Engineer Associate exam on your first attempt.I wish you all the best in your exam!
