README.md

snowflake-notebooks

Unofficial snowflake notebooks created by me

This is an unofficial repo. If you need the latest and greatest, check out the snowflake documentation or Snowflake Developers or Snowflake labs repo.

As of November 2024, I no longer work at Snowflake and will not be updating the notebooks

Notebooks

Basics of Snowflake ML using Diabetes Dataset

Basics of Snowflake ML with Forecasting Chicago Bus Ridership

---

Scaling:

• Big Datasets for showing off Snowpark

• Snowpark with Big Data for Forecasting

• Forecasting M5 Multi Time Series with Nixlta

  • Streamlit Forecasting App

---

Cortex LLM

• Basics of Cortex LLM from Python

• Evaluation of Cortex LLM with LangChain

• Streamlit App for Cortex LLM

• Simple RAG Example using Cortex functions with Streamlit App

• Evaluating RAG using Cortex by Generating Question / Answer Synthetic Pairs and a Model as Judge

• Getting User Feedback using Cortex with Streamlit App

---

ML Functions from Snowflake

• Using Snowflake ML Functions for Forecasting

---

Feature Store:

• Basic Feature Store in Snowflake

• Basics of Snowpark with Forecasting Chicago Bus Ridership including Feature Store

---

ML Tools:

• OpenFE Tutorial / Demo
• Using Embedding Models or Sentence Transformers with Snowflake

---

Misc Notebooks:

• Running Hugging Face Models off Snowpark

• Starter guide on UDFs and UDTFs in Snowflake

• Benchmarking VLLM with Mistral in Container Services

• Model Evaluation and Monitoring with Evidently AI in Snowflake

• ML Monitoring with Snowflake - Summit 2024 Talk

• Training examples using Snowpark ML and Stored Procedures

---

Containers for Snowflake Container Services:

• Mistral running on VLLM and blog post

• Lag LLama Time Series Model and github repo

• H2O LLM Studio

• Fine Tune Llama using Hugging Face TRL - Container Services Setup Notebook, dockerfile, and Training Notebook using TRL

• Use Nemo GuardRails with Snowflake Cortex - based on this example, see folder

---

Model Demos for Snowflake Container Notebooks:

• Llama on VLLM in Container Notebooks

• LLama 3.2 Vision Models on COntainer Notebooks

---

Using the Notebooks:

For connecting to snowflake, the notebooks use a creds.json file. You will want to create this JSON file with the following structure. My newer notebook also include code for using a private key (I typically use private key and 2 factor).

{
    "account":"MY SNOWFLAKE ACCOUNT",
    "user": "MY USER",
    "password":"MY PASSWORD",
    "warehouse":"MY WH",
}