dandelin / ViLT

Code for the ICML 2021 (long talk) paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"
Apache License 2.0
1.41k stars 208 forks source link

ViLT training time #64

Open xii-rao opened 2 years ago

xii-rao commented 2 years ago

Hi

Thanks for your impressive works, which helps me a lot to solve my tasks. Due to the surprised outcome we obtained by fine-tuning the ViLT, we plan to train one for our special case. So, would you mind let me know how long did you train the ViLT with 64 V100 GPUs? So that we can know if it's worth to train. Much appreciate if someone can reply.

S-Moer commented 2 years ago

3days