How to pretrain zero-shot dataset?

microsoft / GLIP

Grounded Language-Image Pre-training

MIT License

2.23k stars 192 forks source link

How to pretrain zero-shot dataset? #116

Open imzhaoruijie opened 1 year ago

imzhaoruijie commented 1 year ago

How can we use word vectors in pre-training? Is it the PREDEFINED_TEXT: odinw/pothole/category_description.json mentioned in the config file? How should I get this json file?

Youskrpig commented 1 year ago

same question

imaginistLi commented 1 year ago

same question

zilong69 commented 1 year ago

same question