Large language models

Dive into the powerful PyTorch framework, an open-source deep learning library that facilitates seamless development and deployment of neural networks. Participants will gain hands-on experience with PyTorch's dynamic computational graph and learn to build, train, and optimize deep learning models efficiently. From basic concepts to advanced techniques, this workshop provides a comprehensive understanding of PyTorch's capabilities for implementing cutting-edge deep learning solutions.

Uncover the secrets of processing textual data with a focus on attention mechanisms. Participants will explore how attention mechanisms enhance the ability of neural networks to process sequential information effectively. Through practical exercises, attendees will learn to implement attention mechanisms in text-based applications, improving model performance and interpretability.

Delve into the Hugging Face Library, a go-to resource for natural language processing enthusiasts and deep learning practitioners. This workshop will guide participants through the extensive functionalities offered by the Hugging Face Transformers library, which provides easy access to pre-trained models for various natural language understanding tasks. Attendees will learn to seamlessly integrate state-of-the-art models, such as BERT and GPT, into their projects, and explore the library's tools for fine-tuning models on custom datasets. From text classification to language generation, this session empowers participants to harness the capabilities of Hugging Face for efficient and effective natural language processing workflows.

Discover the versatility of the einops library and its role in simplifying tensor manipulations. In this workshop, participants will explore how einops can be leveraged to implement attention mechanisms efficiently. Through hands-on coding sessions, attendees will gain practical insights into using einops for reshaping tensors and seamlessly integrating attention mechanisms into their deep learning models.

In this session we will provide students with an overview of how vision language models work. We will then demonstrate the training process of several models, including CLIP, COCA, BLIP, and Flamingo. Following this, we will showcase how to utilize some of these models in practical applications.

Gain a high-level perspective on Large Language Models (LLMs) and their applications. This segment provides an overview of prominent LLMs, such as GPT-3, and delves into the evaluation mechanisms used to assess their performance. Participants will understand the key challenges and considerations in evaluating language models, ensuring a well-rounded understanding of LLMs' capabilities and limitations.

Unlock the power of seamless experimentation with Weights & Biases (wandb), your go-to platform for effortless data logging and analysis. With wandb's intuitive interface and robust logging features, you can track every aspect of your machine learning experiments with ease. From model performance metrics to hyperparameter configurations, wandb empowers you to log, visualize, and compare data in real-time, providing valuable insights to drive your research forward. Whether you're fine-tuning models, exploring new architectures, or collaborating with teammates, wandb simplifies the process, allowing you to focus on what matters most – building smarter, more effective models. Join the thousands of researchers and developers who rely on wandb to streamline their experimentation workflow and accelerate innovation. Start logging your data with wandb today and unleash the full potential of your machine learning projects.

Take a deep dive into loading and working with Large Language Models (LLMs) locally. Participants will learn the intricacies of model deployment and explore practical applications, focusing on retrieval-augmented generation. Through hands-on exercises, attendees will gain the skills to integrate LLMs into their projects, leveraging their power for tasks that require a combination of information retrieval and language generation.