Weihan Wang

你好，我有一个问题

1

为什么tfidf，doc2vec你都做了stacking，而word2vec没有呢，如果加入lda的特征需要进行stacking吗

我邮箱是 [email protected]

Training on a larger dataset

1

Hello, the author, great job! I found that Dino performs very well on the coco dataset. But if I want to transfer the model to a dataset with thousands or...

pretraining task

3

Hello, the author, great work! I'm curious whether you have tried to add Image Text Contrast Learning in the pretraining task? Because in the ALBEF paper, they reported that the...

Information leakage occurs in snli-ve training

2

The main difference between VE and TE task is, the premise in TE in a natural language sentence P~text~, instead of an image premise P~image~. However, in OFA experiment setting,...

hard negative sample selection

2

Hello, the author, great work! I am curious about the process of selecting hard negative samples in the ITC task. Will the samples that are very similar to the positive...

Hello, the author, great work! As time goes by, a lot of image urls in the dataset become invalid. Is there any solution? Could you provide the data arrow?

resume dataloader

6

During the process of using WebDataset for distributed training, my training got interrupted for some reason. Since I have a dataset of up to 2 billion samples, I want to...

Hello Author, Firstly, I'd like to express my admiration for the excellent work you have been doing. It truly stands out in the research community. Recently, some studies have suggested...

Pretraining dataset

1

Thank you for your excellent work. I'm currently training my own CLIP model and have a question. If I use LAION-2B, COYO-700M, and Datacomp datasets simultaneously for training, will it...