Inconsistent results

[xk16111108]

When reproducing the code, using the "Medt" network, you have not encountered the training and test results of the situation is very different

[yunfeihaha]

me too,why?

[xk16111108]

I don't know, in the test, there's still a lump in the upper left corner, and the test is very different from the training

[xk16111108]

[yunfeihaha]

The training and test results of GLAS Dataset are bad

[yunfeihaha]

The train code
python train.py --train_dataset "./Medical-Transformer/dataset/war/train/" --val_dataset "./Medical-Transformer/dataset/war/validation/" --direc './Medical-Transformer/dataset/war/result/' --batch_size 1 --epoch 400 --save_freq 10 --modelname "MedT" --learning_rate 0.001 --imgsize 400 --gray "no"

[xk16111108]

We're about the same. It's not going to work

[rtyasdf]

@yunfeihaha, I leaved a PR, discussing resulting image, check it out, maybe you'll find it useful

[TANGHHH123]

me too,why?

请问您是如何训练的呢？我下载代码后感觉运行不了

[LiaoZihZrong]

The train code python train.py --train_dataset "./Medical-Transformer/dataset/war/train/" --val_dataset "./Medical-Transformer/dataset/war/validation/" --direc './Medical-Transformer/dataset/war/result/' --batch_size 1 --epoch 400 --save_freq 10 --modelname "MedT" --learning_rate 0.001 --imgsize 400 --gray "no"

how did you fix your problem ?

[SpongeBoSquarePants]

When reproducing the code, using the "Medt" network, you have not encountered the training and test results of the situation is very different

Me too.I met the same issue as you.How did you solve it？

[SpongeBoSquarePants]

The issue is disappear after I delete model.eval() in the test.py file.I don't know why.