Synthesized answer
A data scientist might choose to "train transformers from scratch" if they want to learn how to scale these models to multiple GPUs and distributed environments [1]. This approach allows for hands-on learning of the scaling process [1].
The passages do not explicitly state the circumstances under which a data scientist would choose to train transformers from scratch over fine-tuning a pre-trained model. They also do not detail the distinct challenges or benefits associated with this decision specifically when scaling to multiple GPUs and distributed environments, beyond mentioning that it is something one can learn to do [1].
Synthesized from the book passages below. Chat with the book on Feynman for follow-up.
From the book
rs work and how to integrate them in your applications. You'll quickly learn a variety of tasks they can help you solve. Build, debug, and optimize transformer models for core NLP tasks, such as text classification, named entity recognition, and question answering Learn how transformers can be used for cross-lingual transfer learning Apply transformers in real-world scenarios where labeled data is scarce Make transformer models efficient for deployment using techniques such as distillation, pruning, and quantization Train transformers from scratch and learn how to scale to multiple GPUs and…
Description: Since their introduction in 2017, transformers have quickly become the dominant architecture for achieving state-of-the-art results on a variety of natural language processing tasks. If you're a data scientist or coder, this practical book -now revised in full color- shows you how to train and scale these large models using Hugging Face Transformers, a Python-based deep learning library. Transformers have been used to write realistic news stories, improve Google Search queries, and even create chatbots that tell corny jokes. In this guide, authors Lewis Tunstall, Leandro von…
Categories: Computers Pages: 409 Snippet: In this guide, authors Lewis Tunstall, Leandro von Werra, and Thomas Wolf, among the creators of Hugging Face Transformers, use a hands-on approach to teach you how transformers work and how to integrate them in your applications.
Title: Natural Language Processing with Transformers, Revised Edition by Lewis Tunstall, Leandro von Werra, Thomas Wolf
More questions about this book
- The text states transformers have become the "dominant architecture" since 2017. What fundamental limitations or inherent challenges of prior NLP models did the transformer architecture likely overcome to achieve such rapid and widespread dominance across diverse tasks?
- How does the ability to perform "cross-lingual transfer learning" and operate effectively in "scenarios where labeled data is scarce" fundamentally change the accessibility and application of state-of-the-art NLP, especially for languages or domains previously underserved?
- Imagine you are tasked with deploying a transformer model to a low-resource edge device. Explain the specific trade-offs a data scientist must consider when applying "distillation, pruning, and quantization," and how these techniques might alter the model's core functionality or performance.
- The guide emphasizes learning "how transformers work" alongside "how to integrate them in your applications." Why is it critical for a data scientist or coder to understand the underlying mechanics rather than just treating the Hugging Face Transformers library as a black box?