A huggingface transformers implementation of Transformer Memory as a Differentiable Search Index
A huggingface transformers implementation of Transformer Memory as a Differentiable Search Index, Yi Tay, Vinh Q. Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, Tal Schuster, William W. Cohen, Donald Metzler
Requirements: python=3.8
transformers=4.17.0
datasets=1.18.3
wandb
Note: This is not the official repository.
Goal of this repository
Reproduce the results of DSI Large, Naive String Docid, NQ10K. According to Table 3 in the original paper, we should have [email protected]=0.347
,[email protected]=0.605
Step1: Create NQ10K training (indexing) and validation datasets
cd data/NQ
python3 create_NQ_train_vali.py
Step2: Run training script
cd back to the root directory and run: