cesc-park / attend2u

🖼️ Attend to You: Personalized Image Captioning with Context Sequence Memory Networks. In CVPR, 2017. Expanded : Towards Personalized Image Captioning via Multimodal Memory Networks. In IEEE TPAMI, 2018.
MIT License
209 stars 57 forks source link

Doubt about steps and InstaPic dataset size #16

Open adityap27 opened 3 years ago

adityap27 commented 3 years ago

Do we need to train for 5 lac. steps as specified in code ? or for just 20 epochs as per paper ?

Also, earlier I downloaded InstaPic data set, which only had 1.03M samples instead of 1.1M. why ?

Hoping to get a reply soon. Thanks :)