Comment Comment by donsupreme > Unlike GPT, which only uses encoder blocks, or BERT, which only uses decoder blocksI think you meant the reverse, autoregressive model like GPT uses only the decoder while autoencoder like BERT only uses the encoder. parent donsupreme 3 years ago
I think you meant the reverse, autoregressive model like GPT uses only the decoder while autoencoder like BERT only uses the encoder.