Natural Language Processing with Transformers, Revised Edition

Index

A

• absolute positional representations, Positional Embeddings
• abstractive QA, Going Beyond Extractive QA
• abstractive summaries, The CNN/DailyMail Dataset
• Accelerate library
  • about, Hugging Face Accelerate
  • as part of Hugging Face ecosystem, The Hugging Face Ecosystem
  • changes to training loop, Defining the Training Loop
  • comparison with Trainer, Defining the Training Loop
  • infrastructure configuration, The Training Run
  • launching training jobs, The Training Run
• Accelerator
  • is_main_process, Defining the Training Loop
  • prepare(), Defining the Training Loop
  • process_index, Defining the Training Loop
• accuracy metric, Defining the performance metrics, Conclusion, Creating a Performance Benchmark
• ADAPET method, In-Context and Few-Shot Learning with Prompts
• AI Dungeon, Text Generation
• ALBERT model, The Encoder Branch, Span classification
• Amazon ASIN, Initializing a retriever
• Ameisen, Emmanuel, Creating a Performance Benchmark
• analysis, of pretraining run, Results and Analysis-Results and Analysis
• Apache Arrow, A First Look at Hugging Face Datasets, Memory mapping
• argmax, Loading a Custom Model, Greedy Search Decoding, Tokenizing text for QA, Tokenizing text for QA, Optimizing Inference with ONNX and the ONNX Runtime
• ASR (automatic speech recognition), Speech-to-Text
• attention
  • block local, Sparse Attention
  • causal, The Transformer Architecture
  • dilated, Sparse Attention
  • encoder-decoder, The Decoder
  • global, Sparse Attention
  • linearized, Linearized Attention
  • masked multi-head self-, The Decoder
  • multi-headed, Multi-headed attention ...