Glossary

Selected terms from the official Transformers glossary, rewritten for quick reference.

Adapted for this playbook from the 🤗 Transformers documentation by Hugging Face. Official page: https://huggingface.co/docs/transformers/glossary. Images © Hugging Face (documentation-images) unless noted. This is not a substitute for the upstream docs — verify against the current version.

Covers Hugging Face pages: glossary

Term	Meaning
Attention mask	Marks which tokens are real vs padding
Auto class	Factory that reads config.json and loads the right class
Checkpoint	Saved weights (+ config/preprocessor)
Decoder-only	Causal LM architecture (GPT-style)
Encoder-only	Bidirectional encoder (BERT-style)
Encoder–decoder	Seq2seq (T5, Whisper, BART)
Feature extractor	Audio (and some vision) preprocessor
Finetuning	Updating pretrained weights on downstream data
Hub	huggingface.co model/dataset/space hosting
Hidden state	Layer activations
KV cache	Cached keys/values for faster decoding
Pipeline	High-level inference API
Processor	Multimodal preprocessor bundle
Tokenizer	Text ↔ token ids
Trainer	High-level training loop

Full glossary: Hugging Face glossary.

What to do next

Models timeline.