AI Examples

The UFIT Research Computing AI Support Team maintains a suite of examples for computer vision and natural language processing tasks using the AI software stacks on HiPerGator. Users may copy these examples to their own space, add modifications, and follow the instructions to run these jobs on HiPerGator. Each example has a readme file with additional information in its directory.

Use https://support.rc.ufl.edu to submit ticket if you need help with the examples or have any AI questions.

Research Computing Github

We have examples available on the UF Research Computing Github.

Name	Categories	Description
Multinode Pytorch Launch	distributed compute	Launch a pytorch script for multiple GPUs on multiple nodes.
Py4AI	Python education	The episodes in this course are derived from work that is Copyrighted © by Software Carpentry. Additional information can be found at: http://software-carpentry.org/.
Understanding PyTorch	Pytorch education	Learn the Pytorch deep learning software by coding linear regression from the neural network perspective
Pytorch Convolutional Neural Network	computer vision	This training module is based off of the Deep Learning with Pytorch: A 60 minute blitz tutorial by Soumith Chintala. It has been updated and customized for running on HiPerGator

Catalog of available examples

Name	Categories	Location on HiPerGator	Author	Date added	Description
Modulus single-GPU and multi-GPU example	PINN	/data/ai/examples/pinn/modulus	Yang Hong	May, 2024	Modulus single-GPU Jupyter-notebook example is for using PINN to approximate the solution of a given PDE and boundary conditions. The multi-GPU example is for using GraphNN accelerates MD simulations to predict the force of each atom in the system.
RAPIDS	Computer Vision	/data/ai/examples/rapids	Dimitri Bourilkov	April, 2024	Examples for accelerated data science with RAPIDS.
Rapids_singlecell	Healthcare and Life Science	/data/ai/examples/rapids_singlecell	Huiwen Ju and Qian Zhao	May, 2024	Rapids-singlecell offers enhanced single-cell data analysis as a near drop-in replacement predominantly for scanpy
Llama	Healthcare and Life Science	/data/ai/examples/llms/llama	Qian Zhao	May, 2024	The Llama Family from Meta includes Llama 2
NVIDIA Clara Parabricks	Healthcare and Life Science	/data/ai/examples/parabricks	Huiwen Ju and Qian Zhao	May, 2024	NVIDIA Clara Parabricks is a powerful genomics analysis software suite that leverages accelerated computing to process data efficiently.
TensorFlow+Keras Convolutional Neural Net	Computer Vision	/data/ai/examples/image/tensorflow-bootcamp	Dimitri Bourilkov	July, 2021	TensorFlow+Keras Convolutional Neural Net with NGC container
Object Detection Tutorial	Computer Vision	/data/ai/examples/image/4.Object_Detection_Tutorial	Yunchao Yang	Oct, 2022	This example illstrates how to generate a object detection on a video using pretrained models.
OpenCV Intro Tutorial	Computer Vision	/data/ai/examples/image/1.OpenCV_Intro_Tutorial	Yunchao Yang	Oct, 2022	This example illustrates the fundamentals of image processing techniques in opencv.
PyTorch Instance Segmentation Tutorial	Computer Vision	/data/ai/examples/image/5.PyTorch_Instance_Segmentation_Tutorial	Yunchao Yang	Oct, 2022	This example illstrates how to train an instance segementation model using mask r-cnn model
PyTorch Transfer Learning Tutorial	Computer Vision	/data/ai/examples/image/3.PyTorch_Transfer_Learning_Tutorial	Yunchao Yang	Oct, 2022	This example illstrates how to train a convolutional neural network for image classification using transfer learning.
TorchVision Intro Tutorial	Computer Vision	/data/ai/examples/image/2.TorchVision_Transforms_Tutorial	Yunchao Yang	Oct, 2022	This example illustrates the fundamentals of TorchVision in image transformation and augmentation
GraphCNN.ipynb	Graphs	/data/ai/examples/graphs	Eric Stubbs	August, 2021	A torch version of an article on Understanding Graph Convolutional Networks. The example shows how convolution learns from graphs even without training and backpropagation.
MONAI Medical Images Classification Tutorial	Medical Image Processing	/data/ai/examples/image/6.MONAI_Medical_Imaging_Tutorial	Yunchao Yang	Oct, 2022	This example illstrates how to train an medical image classification model using MONAI
AI News GPT	NLP	/data/ai/examples/nlp/ai_news_GPT	Eric Stubbs	March, 2023	See how trained from scratch GPT models can enable knowledge exploration or brainstorming. This example also shows how transfer learning, amount of data, and finetuning affect GPT models.
bertopic_topics.ipynb	NLP	/data/ai/examples/nlp	Eric Stubbs	May, 2021	Apply BERTopic software to examine topics of AI news articles.
embeddings-megatronbert345.ipynb	NLP	/data/ai/examples/nlp	Eric Stubbs	June, 2021	Create vector representations using Nvidia Megatron-BERT-345m.
embeddings-to-bertopics.ipynb	NLP	/data/ai/examples/nlp	Eric Stubbs	July 2021	Use cosine similarity, a similarity matrix, and a centrality matrix to identify key sentences within an article that provide a summary of the main points.
Example Multinode GPT Pretraining	NLP	/data/ai/examples/distributed-compute/pytorch_distributed_exampleGPT	Eric Stubbs	July, 2021	Use the pytorch distributed launch utility to pretrain a GPT language model using multiple nodes.
Improved Megatron GPT Text Generation	NLP	/data/ai/examples/nlp/megatronGPT_text_generation/megatronGPT_text_generation	Eric Stubbs	August, 2022	Megatron code was updated to allow for bulk prompt completion and language generation using a 23 billion parameter Megatron GPT language model trained by UFIT Research Computing.
inferences-megatronbert345.ipynb	NLP	/data/ai/examples/nlp	Eric Stubbs	July 2021
Language Model Inference Apps	NLP	/data/ai/examples/nlp/inference_apps	Eric Stubbs	May, 2022	Explore knowledge in Nvidia Megatron BERT and GPT models using notebooks by creating lists of the most likely predicted words.
MNLI Benchmark with UF Model	NLP	/data/ai/examples/nlp/multinli_benchmark	Eric Stubbs	August, 2022	Apply a 9 billion parameter Megatron BERT language model trained by UFIT Research Computing to an inference benchmark using improved code from Megatron. The model achieved a score of 87% after only 1 epoch of finetuning. State of the art score on NLI is 92%.
NeMo Text Classification	NLP	/data/ai/examples/nlp/nemotextclassify	Eric Stubbs	November, 2021	Train a language model to classify text with a set of labels. Requires labelled data.
Race Benchmark with UF Model	NLP	/data/ai/examples/nlp/race_benchmark_megatron	Eric Stubbs	August, 2022	Apply a 9 billion parameter Megatron BERT language model undertrained on only 400gb of data by UFIT Research Computing to a question answering benchmark using code from Nvidia Megatron-LM. The model achieved a score of 85% on middle school questions and 81% on high school questions after only 1 epoch on the QA training data. State of the art score on RACE is 92%.
summary-extractive.ipynb	NLP	/data/ai/examples/nlp	Eric Stubbs	May 2021