Led by Geoffrey Hinton Simulacrum
The transformer architecture, GPT and BERT model families, embeddings, question answering, and fine-tuning pre-trained models for specific tasks.
Led by Geoffrey Hinton Simulacrum
The question
The problem with RNNs · the attention mechanism · the transformer architecture (input embeddings, multi-headed attention, feed-forward layers, masked attention, output prediction) · what GPT means (Generative Pre-trained Transformer) · the developmen...
Outcome
Demonstrates competence in transformer architecture and gpt.
Sub-units
Led by Geoffrey Hinton Simulacrum
The question
GPT vs BERT (generative vs bidirectional) · BERT architecture (bidirectional encoder, masked language modelling, next sentence prediction) · BERT embeddings · building a question-answering bot with BERT · BERT variants (RoBERTa, DistilBERT) · XLNet (...
Outcome
Demonstrates competence in bert, xlnet and fine-tuning.
Sub-units