hku_diffusion_dllm / reference /code /flm /configs /data /openwebtext-split-reflow.yaml
Ouzhang's picture
Add files using upload-large-folder tool
31dc8dc verified
Raw
History Blame Contribute Delete
215 Bytes
train: reflow-dataset
valid: openwebtext-valid
tokenizer_name_or_path: gpt2
cache_dir: /PATH/TO/TRAIN/DATASET
wrap: True
streaming: False
insert_train_eos: True
insert_valid_eos: True
save_dir: /PATH/TO/SAVE/DATASET