张昱辉Zhang Yuhui

icon indicating a website link https://www.fudan.edu.cn/ icon indicating an email [email protected]

icon company Fdu icon location 2205,songhu avenue,shanghai city,china icon location My name is Zhangyuhui, A junior in the School of Computer Science, Fudan university , majoring in software engineering, enjoys programming

Results 6 comments of 张昱辉Zhang Yuhui

How can I change the scale for training?

when i tried i change var.py class VAR's patch_nums, class VARHF's patch_nums,and arg_util.py pn

How can I change the scale for training?

@iFighting thanks a lot! I am working on retrain var for inpainting research, hope i can get a good result :)

How can I change the scale for training?

still trying    ------------------ Original ------------------ From: ***@***.***>; Date:  Wed, Dec 18, 2024 03:47 PM To: ***@***.***>; Cc:  "张昱辉Zhang ***@***.***>; ***@***.***>; Subject:  Re: [FoundationVision/VAR] How can I change the scale for training?...

is it make sense to try image loss in stage 2?

also i find sometimes used gumbel-softmax(hard=false) will show a better result for train,but it is a bad set for use hard=false? if codebook is limited, will mixed token show more...

is it make sense to try image loss in stage 2?

@jack111331 thx for u reply , I will try what your questioned in serveral weeks due to busy work. the mix token will useful if loss is not only crossentropy...

is it make sense to try image loss in stage 2?

@Longhzzz I use gumbel_softmax for different scales token that mix into the final 16*16 token map(which then through frozen decoder into img ,then use img to apply image loss. I...