a-PyTorch-Tutorial-to-Image-Captioning icon indicating copy to clipboard operation
a-PyTorch-Tutorial-to-Image-Captioning copied to clipboard

Published 20 hours ago • verified publisher icon sgrvinod

Question about the checkpoint

Open Jazz1996 opened this issue 5 years ago • 8 comments

Thanks for your wonderful project! I tried to train the model using your way. Firstly I trained only the decoder with the batch size of 80 without fine-tune for 15 epochs. Then I tried to use fine-tune and recover to the checkpoint, but after I added the checkpoint path and ran it again, the training didn't begin. Can you help me solve this problem? Thanks! 屏幕快照 2019-04-03 下午2 30 32

Jazz1996 avatar Apr 03 '19 06:04 Jazz1996

Hi @Jazz1996 , thank you, does this still happen? If it does, let's try to figure out why and if there's a bug, I'd like to fix it.

When this happened, checkpoint was set to the path of the un-fine-tuned checkpoint and fine_tune_encoder was set to True, correct?

sgrvinod avatar Apr 10 '19 05:04 sgrvinod

Hi @sgrvinod , correct, and the training still doesn't begin.

Jazz1996 avatar Apr 18 '19 02:04 Jazz1996

Hi @Jazz1996 ,Have you solved this problem, I hope I can get your help

feixiangqiqi avatar Mar 26 '22 13:03 feixiangqiqi

Hello, is there any solution to this problem?

aliciaviernes avatar Jun 28 '22 15:06 aliciaviernes

It was a long time, so I can hardly remember this problem now. Sorry I do not have a solution to it.

Jazz1996 avatar Jun 28 '22 16:06 Jazz1996

is there any solution to this problem?

supriamir avatar Sep 29 '22 09:09 supriamir

Yes, check your training loop. Maybe you'll find the range something like (10,10). So that's why it doesn't show any error. Check what value is assigned to start_epoch.

LavanyaGovindaraju avatar Mar 10 '23 23:03 LavanyaGovindaraju

Hello，the epoch you set in your new train should bigger than the one in your last train.

Qiqi224 avatar Mar 18 '24 16:03 Qiqi224