search

DeyvidKochanov-TomTom / kprnet

MIT License
65 stars 12 forks source link

Questions about pre-trained model #5

Open Zhen-ao opened 3 years ago

Zhen-ao commented 3 years ago

hi~Thanks for your work, but I have a question： Is it possible to not load the pre-trained model? Will it be very different from the current result?

DeyvidKochanov-TomTom commented 3 years ago

Good question, as of now I haven't done proper ablation so I can't tell you exactly how much worse it would be - I expect maybe 2 points if you use the Imagenet pre-trained model and worse if you do random initialization also training converges slower. May I ask what is the problem with the pre-trained weights.