recognize-anything icon indicating copy to clipboard operation
recognize-anything copied to clipboard

why the tag number is 40898

Open JacobLee121 opened this issue 1 year ago • 1 comments

for ram++ the label_embed size is [2085798, 512]. since each tag has 51 descriptions, so the tags number is 40898 = 5085798/51. but in the configs of finetune.yaml for ram++ is 4585 tags. why the real tag number in model is 40898 not 4585

JacobLee121 avatar Nov 16 '23 03:11 JacobLee121

The label embedding of RAM++ is [233835,512], in here https://huggingface.co/xinyu1205/recognize-anything-plus-model/blob/main/ram_plus_tag_embedding_class_4585_des_51.pth

xinyu1205 avatar Nov 16 '23 07:11 xinyu1205