Tensor Shape size issue

[mshaheryar91]

Hello,
I am running the code and in second epoch, it says the in and out tensor size issue. The in is 3,256,256 and out is 3,248,258. I tried to use resize function. I have used the following configuration:
batch_size: 8
data_repeat: 80
data_augment: 1
epochs: 1000
lr: 0.00025

Please guide!

[mshaheryar91]

Waiting for kind reply. If someone can guide me on this

[songzijiang]

and how you set the following parameters?
patch_size:
batch_size:
I think you should try the commend settings first.

[mshaheryar91]

patch_size: 256
batch_size: 8
data_repeat: 80
data_augment: 1
epochs: 1000
lr: 0.00025
decays: [250, 400, 450, 475, 500]
gamma: 0.5
log_every: 100
test_every: 1

I have tried it but Cuda got out of memory, so i tried diffreent setting and then it start learning but then in 2nd epoch, it got this error

[songzijiang]

It looks ok, maybe you can offer the whole config file and console output.

[mshaheryar91]

elan_light_x4.yml.txt
For Console output, i will rerun the code and then attach it

[mshaheryar91]

This is first epoch result

[mshaheryar91]

Please find error, its another error now, I am quite surprised why on every iteration it gives new error, related to size, resize, tensor shape. What do you think possible reasons for that?

[mshaheryar91]

may be I found the reason for that , please see attachment (when i rerun the code again and i found that in first epoch it is not training the whole datasets and giving nan value so i think in every epoch it is missing some files but the code is same so why this is happening.

[songzijiang]

I have no idea how to use it in the windows system, maybe you could download the dataset provided in README for a try. Or you can wait for the author‘s response.