承认自己是普通人，我们只要会用就行了。

没有Machine Learning，也没有Deep Learning，让95%的普通码农都会用大模型，至于底层原理，就让那5%的精英去学吧。

这是“Generative AI with Large Language Models”课程的学习笔记，了解大模型项目的生命周期，创建自己的大模型应用。

内容介绍

Transformer Architecture and Inference Techniques

1. Examine the transformer architecture that powers large language models (LLMs).

        探讨驱动大语言模型（LLMs）的Transformer架构。

2. Explore how these models are trained.

        探讨这些模型的训练方法。

3. Understand the compute resources required to develop these powerful LLMs.

        了解开发这些强大LLMs所需的计算资源。

4. Learn about in-context learning and how to guide the model at inference time with prompt engineering.

        学习上下文学习以及如何通过提示工程在推理时引导模型。

5. Tune the most important generation parameters of LLMs to optimize model output.

        调整LLMs的最重要生成参数以优化模型输出。

Instruction Fine Tuning

1. Explore options for adapting pre-trained models to specific tasks and datasets via instruction fine tuning.

        探讨通过指令微调将预训练模型适应于特定任务和数据集的选项。

2. Understand the process and benefits of instruction fine tuning.

        了解指令微调的过程和优势。

Aligning LLMs with Human Values and Reinforcement Learning from Human Feedback (RLHF)

1. See how to align the output of language models with human values to increase helpfulness and decrease potential harm and toxicity.

        学习如何调整语言模型的输出，使其符合人类价值观，以提高有用性并减少潜在的伤害和有害内容。

2. Learn about reinforcement learning from human feedback (RLHF).

        学习从人类反馈中进行强化学习（RLHF）。

3. Build a reward model classifier to label model responses as either toxic or non-toxic.

        构建一个奖励模型分类器，将模型响应标记为有害或无害。