Pretrained weight tags for lucid.models.gpt2_large_lm.
GPT-2 Large trunk + tied lm_head over the 50 257-token BPE vocab.
Attributes
WEBTEXTWeightEntryWebText pre-training checkpoint (trunk + LM head) sourced from
transformers/gpt2-large.DEFAULTWeightEntryAlias for
WEBTEXT.Notes
Reference: Radford, Wu, Child, Luan, Amodei, Sutskever, "Language Models are Unsupervised Multitask Learners", OpenAI 2019.
Examples
>>> from lucid.models import gpt2_large_lm
>>> model = gpt2_large_lm(pretrained=True).eval()