Pretrained weight tags for lucid.models.gpt2_medium.
Radford et al. 2019 24-layer trunk (H=1024, A=16, 355 M params).
Attributes
WEBTEXTWeightEntryWebText pre-training checkpoint sourced from
transformers/gpt2-medium.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_medium
>>> model = gpt2_medium(pretrained=True).eval()