Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
We assume that the folder models/megatron_gpt2 contains
megatron_gpt2_345m_v0_0.zip and that the command is run from that folder:
python3 $PATH_TO_TRANSFORMERS/models/megatron_gpt2/convert_megatron_gpt2_checkpoint.py megatron_gpt2_345m_v0_0.zip
MegatronGPT2 architecture is the same as OpenAI GPT-2 .