search

zcfinal / MM-Rec

Source codes for paper "MM-Rec: Visiolinguistic Model Empowered Multimodal News Recommendation".
22 stars 3 forks source link

reproduce #5

Open xylovezxy opened 3 months ago

xylovezxy commented 3 months ago

HI，excuse me, What kind of dataset can I use to reproduce this paper,How about movielens dataset

zcfinal commented 3 months ago

I am not sure whether the movie dataset maintains the same results. However, you can try this dataset: https://drive.usercontent.google.com/download?id=1gx0OzN7qSuyRlvN0cfVUjB4tmoKvvQk1&export=download&authuser=0 This dataset is from Why Do We Click: Visual Impression-aware News Recommendation, which was crawled according to MIND dataset.

xylovezxy commented 3 months ago

Thank you very much. I see that you are using eight GPUs, so what is the minimum requirement for GPUs for this project.

zcfinal commented 3 months ago

We used 8 V100 GPUs for data parallelism to speed up training. One V100 GPU is able to support training, but the training is slower.

xylovezxy commented 3 months ago

Thank you very much