train customer data - Githubissues

dandelin / ViLT

Code for the ICML 2021 (long talk) paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"

Apache License 2.0

1.36k stars 209 forks source link

Open guanhdrmq opened 2 years ago

guanhdrmq commented 2 years ago

Hi, if I train a new dataset, do I still use the data preparation script? For example, MM-IMDb and UPMC Food-101 dataset. Thank you very much.

Ellyuca commented 1 year ago

Hi @guanhdrmq. Did you manage to figure out how to prepare data from new datasets? Thanks in advance.