search

huochaitiantang / pytorch-deep-image-matting

Pytorch implementation of deep image matting
294 stars 71 forks source link

About learning rate setting #31

Open myCigar opened 4 years ago

myCigar commented 4 years ago

Hello @huochaitiantang

Recently, I retrained the model with your code. The learning rate in the train.sh is 0.00001, and the default value in the train.py is 0.001. Which learning rate is the one you are using for your current training model? Thanks !

huochaitiantang commented 4 years ago

lr=1e-5 as the paper.

myCigar commented 4 years ago

Thanks.

hejm37 commented 4 years ago

Hi, @huochaitiantang

I am wondering if the training is stable (replicable)? I ran the experiment using the default setting provide by train.sh (except that I change the batchsize to 25), but I got sad=66.7 after epoch 25 and the best sad=60.7 after epoch 5. Do you have any ideas? Thanks!

huochaitiantang commented 4 years ago

Current training is sensitive to batch size and the large batch will lead to lower performance. You could get a reasonable result with batch size = 1.

hejm37 commented 4 years ago

Thanks! I will try a few more experiments (with batch size=1) to see if I could get close to sad=54.42.

hejm37 commented 4 years ago

Hi, @huochaitiantang

I tried batch size = 1 as you said and got reasonable results from 2 out of 3 experiments. However, I found that the best performance was from Epoch 11 and Epoch 17 respectively and then the sad went up again (more accurately, it was fluctuating at a range 55~70). Were you having the same situation in your experiment? Thanks!

huochaitiantang commented 4 years ago

In my experiment, the sad metric fluctuates at a range 54~64 after the 3rd epoch.

hejm37 commented 4 years ago

Thanks for your response! I found the same in my experiment.

kfeng123 commented 4 years ago

Hi, thank you for your excellent work! My question is: why the performance gets worse by increasing the batch size? Does this phenomenon documented in the literature? I am very confusing.