requirements.txt

numpy==2.0.1
torch==2.4.0
einops==0.7.0 # used in implementation of dual attention / relational attention
huggingface-hub==0.23.4 # for loading pre-trained checkpoints
bertviz==1.4.0 # for visualization
tiktoken==0.7.0 # tokenizer used for LM experiments
gradio==4.40.0 # for inference/visualization apps
dual_attention==0.0.4
tqdm==4.66.5