IBM-Granite-Hackathon

Learning Granite for the Sprint

This page is the sprint onboarding hub for Granite.

---

What Is IBM Granite?

IBM Granite models are designed for tasks such as code generation, natural language understanding, summarization, and more. They are available through IBM’s watsonx.ai platform and as open-source models on Ollama and Hugging Face.

Key properties:

• Open and transparent foundation models
• Enterprise-oriented safety and governance focus
• Strong fit for coding and agentic tasks in this sprint context

---

Sprint Learning Path (Recommended)

1. Start with one recipe from the Granite agent cookbook.
2. Reproduce it locally and document what you changed.
3. Choose your direction: Granite agent workflow.
4. Define an MVP you can demo by April 18.

---

Official Documentation & Getting Started

Resource	Link
IBM Granite product page	https://www.ibm.com/granite
watsonx.ai platform	https://www.ibm.com/products/watsonx-ai
IBM Granite on Ollama	https://ollama.com/ibm/granite4
IBM Granite on Hugging Face	https://huggingface.co/ibm-granite
Granite model documentation	https://www.ibm.com/granite/docs
Granite Agent Cookbook	https://github.com/ibm-granite-community/granite-agent-cookbook

---

Sprint Project Tracks

Granite Agents

Build agentic workflows for coding, documentation, or analysis tasks.

---

Granite Model Families

Granite Language Models

General-purpose large language models for text tasks (summarization, classification, Q&A, generation).

• ibm-granite/granite-4.0-micro — 3B parameter instruction-tuned model
• ibm-granite/granite-4.0-1b — 1B parameter instruction-tuned model

Granite Guardian

A model for safety and trust evaluation.

• ibm-granite/granite-guardian-3.3-8b

---

Hands-On Tutorials & Notebooks

Tutorial	Link
Getting started with watsonx.ai	https://developer.ibm.com/tutorials/awb-getting-started-watsonx-ai/
Prompt engineering basics	https://www.ibm.com/topics/prompt-engineering
Granite snack cookbook	https://github.com/ibm-granite-community/granite-snack-cookbook
IBM Granite community GitHub	https://github.com/ibm-granite-community
Granite agent cookbook	https://github.com/ibm-granite-community/granite-agent-cookbook

---

Key Concepts to Learn

1. Prompt Engineering — How to craft effective prompts to get the best output from a model.
   • IBM guide to prompt engineering
2. Retrieval-Augmented Generation (RAG) — Combine a language model with your own data.
   • What is RAG?
3. Fine-tuning — Adapt a pre-trained model to a specific domain or task.
   • Fine-tuning overview
4. LangChain / LlamaIndex with Granite — Popular frameworks for building LLM applications.
   • LangChain docs
   • LlamaIndex docs

---

APIs & SDKs

Tool	Link
watsonx Python SDK (ibm-watsonx-ai)	https://ibm.github.io/watsonx-ai-python-sdk/
IBM Cloud API docs	https://cloud.ibm.com/apidocs/watsonx-ai
OpenAI-compatible API on watsonx	https://www.ibm.com/docs/en/watsonx/saas?topic=apis-text-generation

---

Community & Support

• IBM Granite GitHub organization
• IBM Granite community
• IBM Developer community
• Ask questions during Office Hours with BJ Hargrave from IBM!

---

Recommended Week Plan

Before Kickoff
├── Read the IBM Granite product page
├── Skim Granite agent cookbook recipes
└── Pick a tentative project track

Kickoff Day (Sunday, April 12)
├── Attend BJ Hargrave's workshop
├── Confirm tooling setup
└── Form teams and choose scoped goals

Sprint Week (April 13–18)
├── Build MVP and test daily
├── Attend office hours when blocked
├── Prepare one meaningful contribution artifact
└── Finalize slides + demo for closing session

---

Back to Home