A Comprehensive Training on HPE’s AI Solution Offerings: Build, Deploy and Accelerate

Overview of SmartSim: AI integration with HPC simulations.

• Benefits for researchers, engineers: improved communication, efficiency.
• Key functionalities: HPC workload automation, TensorFlow/PyTorch/ONNX support.

Library Design

• SmartSim components: SmartSim (infrastructure) and SmartRedis (client).
• Individual features: HPC job management, Redis deployment, ensembles.
• SmartRedis capabilities: tensor support, model evaluation, distributed placement.

Use Cases of SmartSim

• Online Training
• Using neural networks as surrogate models.
• Real-time predictions reduced computational load.
• Online Inference
• Deploying trained models for real-time inference.
• Orchestrator's role in online inference.
• Standard and co-located modes for database coupling.
• Online Analysis
• Real-time visualization and analysis of simulation data.

Hands-on Lab: Familiarization and Applications of SmartSim

• Introducing Anomaly Detection: Unveiling insights hidden in data.
• Harnessing the power of Isolation Forests: Detecting anomalies with trees in the forest.
• Autoencoders: Unsupervised anomaly detection with neural networks.

Hands-on Labs: Design of Anomaly Detector using Autoencoder Architecture

Computer Vision:

• Dive into the world of Computer Vision.
• From Pixels to Insights: How AI interprets and analyzes images.
• Convolutional Neural Networks
• Image Classification and Object Detection using YOLO.

Natural Language Processing (NLP):

• Introduction to NLP workloads.
• Text Preprocessing
• Text Classification
• LSTM for NLP

Hands-on Labs: Design of AI Applications

Generative AI:

• Discriminative AI Verses Generative AI
• Sequence to Sequence Models
• Transformers
• GPT and BERT
• Fine tuning of LLMs

Hands-on Labs: Design of Generative AI model fine tuning and Object Detection using GPU.

Pipeline Overview:

• Understanding the concept of ML pipelines and their importance.
• Components of a pipeline: Data processing, model training, and serving.
• Creating an efficient and reproducible workflow.
• Reusable Model Servers and Language Wrappers:

Data Management and Versioning using Pachyderm:

• Repos in Pachyderm
• Pipelines in Pachyderm
• Data Versioning in Pachyderm

Seldon Core for Model Deployment:

• Overview of Seldon Core as a machine learning deployment solution.
• Creating, configuring, and deploying a machine learning deployment using Seldon.
• Understanding the structure of a machine learning deployment graph.

Pachyderm Determined Seldon Integration

Hands-on Labs: Pachyderm Seldon Design and Integration

Drift Detection with Alibi-Detect

• Introduction to Drift Detection
• Types of Drift Detected by Alibi-Detect
• Using Alibi-Detect for Drift Detection