Graphs and Transformers - Exploring the Next AI Frontiers
Explore the evolution of deep learning models, focusing on graphs and transformers as fundamental building blocks in modern AI architectures for natural language, protein research, and image generation.
Accelerating Data Intensive ML Applications with the Graphcore IPU
Explore how Graphcore's Intelligence Processing Unit accelerates AI development, focusing on BERT optimization and implementation using Spell's Cloud MLOps platform and HuggingFace tools.
A User Guide to Transformers
Explore practical applications of transformers with minimal examples for various scenarios, enhancing your understanding of this innovative data analysis tool.
Fine-Tuning Foundation Models with Azure Machine Learning Studio
Fine-tune a BERT foundation model using Azure Machine Learning Studio for enhanced natural language processing capabilities.
How Well Do Sparse Models Transfer? - Exploring Transfer Performance in Computer Vision and NLP
Explore sparse models' transfer performance and efficiency gains in image and language tasks, outperforming dense models even at high sparsities.
Linear Structures in Transformer Embedding Spaces
Explore linear structures in Transformer embedding spaces, examining anisotropy, BERT's next sentence prediction, layer behavior, and multi-head attention modules.
Introduction to Language Models - LLMs, Prompt Engineering, and Architectures
Explore language models, covering generative AI, LLMs, encoders, prompt engineering, and transformer architecture. Gain insights into embeddings, attention mechanisms, and model limitations.
From Zero to SOTA: Fine-Tuning BERT Using Hugging Face for World-Class Training Performance
Discover how to fine-tune BERT using Hugging Face and Graphcore's IPU for state-of-the-art NLP performance. Learn to accelerate AI development, optimize models, and leverage cloud platforms for efficient machine learning workflows.
The Lottery Ticket Hypothesis for Gigantic Pre-Trained Models
Explore the lottery ticket hypothesis in large pre-trained models, uncovering efficient subnetworks for transfer learning in NLP and computer vision tasks.
Encoder-Only Transformers: Understanding BERT and RAG Architecture
Discover how encoder-only transformers power RAG, sentiment analysis, and classification through clear explanations of word embedding, positional encoding, and attention mechanisms.
The Evolution of Domain Transfer in the Era of Pre-trained Language Models
Explore the complex relationship between pre-trained language models and domain transfer, examining how pre-training affects model understanding and pattern recognition across different contexts.