晶晶 苗

GPT adopts the strategy of large-scale unsupervised pre-training and learns rich language knowledge from massive corpus of texts. This makes the GPT model perform well in various NLP tasks and is more efficient than traditional classification and clustering methods.