Decoder Layers

Components in sequence-to-sequence models (e.g., transformers) responsible for generating output sequences from encoded inputs. Each layer typically contains self-attention and feed-forward sub-layers, processing tokens incrementally to produce the final output.

Related Concepts

• Transformer
• Encoder-Decoder Architecture
• Automatic Speech Recognition
• Whisper

Applications

• Used in Whisper’s whisper-large-v3-turbo model for real-time ASR, as demonstrated in Fahd Mirza - getting Whisper working on Google Colab.

Backlink: 2026 04 14 Fahd Mirza getting Whisper working on Google Colab