From ab06f78eca9df4bf2d55ee662dc2997bd41bed79 Mon Sep 17 00:00:00 2001 From: Ravindra Patil <162299507+ravipatil33@users.noreply.github.com> Date: Tue, 26 Nov 2024 14:43:59 +0530 Subject: [PATCH] Getting started with RHEL AI Signed-off-by: Ravindra Patil --- knowledge/Desktop/attribution.txt | 5 ++ knowledge/Desktop/qna.yaml | 114 ++++++++++++++++++++++++++++++ 2 files changed, 119 insertions(+) create mode 100644 knowledge/Desktop/attribution.txt create mode 100644 knowledge/Desktop/qna.yaml diff --git a/knowledge/Desktop/attribution.txt b/knowledge/Desktop/attribution.txt new file mode 100644 index 0000000..6246e71 --- /dev/null +++ b/knowledge/Desktop/attribution.txt @@ -0,0 +1,5 @@ +Title of work: Red Hat Enterprise Linux AI1.2 Getting Started +Link to work: https://docs.redhat.com/en/documentation/red_hat_enterprise_linux_ai/1.2/pdf/getting_started/Red_Hat_Enterprise_Linux_AI-1.2-Getting_Started-en-US.pdf +Revision: 1.2 +License of the work: Apache-2 +Creator names: Red Hat Pvt Ltd. diff --git a/knowledge/Desktop/qna.yaml b/knowledge/Desktop/qna.yaml new file mode 100644 index 0000000..bbbd3e8 --- /dev/null +++ b/knowledge/Desktop/qna.yaml @@ -0,0 +1,114 @@ +created_by: ravipatil33 +version: 3 +domain: Technology +document_outline: Includes most basic introduction to RHEL AI Product +seed_examples: + - context: >- + Red Hat Enterprise Linux AI is a platform that allows you to develop + enterprise applications on open source Large Language Models (LLMs). RHEL + AI is built from the Red Hat InstructLab open source project. + questions_and_answers: + - question: What is RHEL AI ? + answer: >- + Red Hat Enterprise Linux AI is a platform that allows you to develop + enterprise applications on open source Large Language Models (LLMs). + RHEL AI is built from the Red Hat InstructLab open source project + - question: What are the components of RHEL AI ? + answer: >- + RHEL AI components includes : Granite family models, InstructLab + upstream project, RHEL Image Mode and Red Hat Enterprise support. + - question: What are the differences in RHEL AI and InstructLab ? + answer: >- + InstructLab is an open source AI project that facilitates + contributions to Large Language Models (LLMs). RHEL AI takes the + foundation of the InstructLab project and builds an enterprise + platform for LLM integration on applications. + - context: >- + Skill and knowledge are the types of data that you can add to the taxonomy + tree. You can then use these types to create a custom LLM model fine-tuned + with your own data. + questions_and_answers: + - question: >- + What are the two types of data which can be added into the taxonomy + data ? + answer: >- + Skill and knowledge are the types of data that you can add to the + taxonomy tree. + - question: What is Knowledge ? + answer: >- + Knowledge for an AI model consists of data and facts. When creating + knowledge sets for a model, you are providing it with additional data + and information so the model can answer questions more accurately. + - question: What is skill ? + answer: >- + A skill is a capability domain that intends to train the AI model on + submitted information. When you make a skill, you are teaching the + model how to do a task. + - context: >- + Red Hat Enterprise Linux AI contains various distinct features and + consists of the following components. Bootable Red Hat Enterprise Linux + with InstructLab, InstructLab model alignment and Open source licensed + Granite models. + questions_and_answers: + - question: What are the supported installation methods in RHEL AI ? + answer: >- + You can install RHEL AI and deploy the InstructLab tooling using a + bootable RHEL container image provided by Red Hat. The current + supported installation methods for this image are on Amazon Web + Services (AWS), IBM Cloud, and bare-metal machines with + - question: What is included in RHEL AI Image ? + answer: >- + This RHEL AI image includes InstructLab, RHEL 9.4, and various + inference and training software, including vLLM and DeepSpeed. + - question: What is LAB ? + answer: >- + InstructLab uses a novel approach to LLM fine-tuning called LAB + (Large-Scale Alignment for ChatBots). The LAB method uses a + taxonomy-based system that implements high-quality synthetic data + generation (SDG) and multi-phase training. + - context: >- + Various hardware accelerators require different requirements for serving + and inferencing as well as installing, generating and training the + granite-7b-starter model on Red Hat Enterprise Linux AI. + questions_and_answers: + - question: What is end-to-end workflow in RHEL AI ? + answer: >- + The end-to-end workflow includes: synthetic data generation (SDG), + training, and evaluating a custom Granite model. + - question: >- + What is the minimum GPU memory required for end-to-end workflow on + Nvidia GPU ? + answer: 'On NVIDIA GPU, the mimimum possible GPU memory is 320 GB. ' + - question: >- + What is the mimimum possible GPU memory for inferencing on Nvidia GPU + ? + answer: 'On NVIDIA GPU, the mimimum possible GPU memory is 24 GB. ' + - context: >- + This glossary defines common terms for Red Hat Enterprise Linux AI : + InstructLab, Large Language Models, Synthetic Data Generation, + Fine-tuning, LAB, Multi-phase training, Serving, PyTorch, Granite, + Taxonomy, PyTorch, vLLM, FSDP etc. + questions_and_answers: + - question: What is Taxonomy ? + answer: >- + The LAB method is driven by taxonomies, an information classification + method. On RHEL AI, you can customize a taxonomy tree that enables you + to create models fine-tuned with your own data. + - question: What is Granite ? + answer: >- + An open source (Apache 2.0) Large Language Model trained by IBM. On + RHEL AI you can download + + the granite-7b-starter model as a base LLM for customizing. + - question: What are the Python Libraries used in RHEL AI ? + answer: >- + PyTorch : An optimized tensor library for deep learning on GPUs and + CPUs, vLLM : A memory-efficient inference and serving engine library + for LLMs and FSDP : An acronym for Fully Shared Data Parallels used + for training and fine-tuning. +document: + repo: https://github.com/ravipatil33/taxonomy-knowledge-docs + commit: c833bfa7eb6cc2b2cd893daaf87ec63b6e596e49 + patterns: + - RHELAI known_issues-20241126T090854692.md + - rhelai1.2_gs-20241126T090854692.md