I have tried many combinations of learning rate and decay still i am getting the same

Out of memory error about rcan HOT 12 CLOSED

yulunzhang commented on July 28, 2024

Out of memory error

from rcan.

Comments (12)

yulunzhang commented on July 28, 2024

It seems that you cannot even begin to train. What's your GPU? The GPU memory used to run the training code is about 7.5--8Gb. Maybe your GPU memory is not enough.

from rcan.

arun121cs commented on July 28, 2024

Thanks. Can you suggest hyper-parameters for 1GB GPU memory? Or else is there pre-trained model which i can test on my own images?

from rcan.

yulunzhang commented on July 28, 2024

First, it is really hard to run experiments with 1 Gb GPU memory. Try to use small input patch size, like 24. Use small batch size, like 1.
The pretrained models are available in Dropbox. Please see the readme file.

from rcan.

arun121cs commented on July 28, 2024

Thanks a lot.

from rcan.

arun121cs commented on July 28, 2024

One last question, do I need to prepare data for testing even if I use the data given by you?

…

On Wed, Jul 25, 2018, 1:51 PM Yulun Zhang ***@***.***> wrote: First, it is really hard to run experiments with 1 Gb GPU memory. Try to use small input patch size, like 24. Use small batch size, like 1. The pretrained models are available in Dropbox. Please see the readme file. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Ahr4U2pN1-yOGmlPwFBue0NbArWrVXf5ks5uKCqDgaJpZM4Vfh9M> .

from rcan.

yulunzhang commented on July 28, 2024

Download original standard testing data and run the matlab file to produce testing LR images. Then run testing code. More details are available in readme file.

from rcan.

arun121cs commented on July 28, 2024

Hello Sir, I am trying to test the model using B100 dataset and i am not able to run the scripts you have mentioned in readme since it gives error that no file found. I used my own path but still i am not able to run it. Kindly tell me the changes in options.py. My path to the B100 folder is /Desktop/Projects/RCAN/RCAN_TestCode/testset. Kindly tell me where to make changes and what they should be. I tried a lot of things but /model/RCAN_BIX2.pt --test_only --save_results --chop --save 'RCAN' --testpath /media/yulun/Disk10T/datasets/super-resolution/LRBI --testset Set5# RCAN_BIX3 is certainly not working.

…

On Wed, Jul 25, 2018, 12:35 PM Yulun Zhang ***@***.***> wrote: It seems that you cannot even begin to train. What's your GPU? The GPU memory used to run the training code is about 7.5--8Gb. Maybe your GPU memory is not enough. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Ahr4U4ecweAm-xeXUdjoPjpypVbXCAjkks5uKBjXgaJpZM4Vfh9M> .

from rcan.

arun121cs commented on July 28, 2024

Hey man, Thanks a lot. This is working for me. I know i asked you silly questions so forgive me. If you want, just tell me one last thing, what is the significance of using degradation in super resolution part. I mean why we need to run two scripts to get output? Are both the outputs are different?

…

On Wed, Jul 25, 2018, 2:10 PM Yulun Zhang ***@***.***> wrote: Download original standard testing data and run the matlab file to produce testing LR images. Then run testing code. More details are available in readme file. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Ahr4UzfiOTOAysQNBj0TzuNYq7HGTMTbks5uKC7sgaJpZM4Vfh9M> .

from rcan.

yulunzhang commented on July 28, 2024

Hi,
The testing pipeline for the two degradation models keeps the same.
I added '--degradation BD' for BD degradation for easier file naming. For example, an LR image is named as 'bird_LRBD_x2.png'. After super-resoltuion, we name the output as 'bird_RCAN_x2.png'.

If you don't use it, it's fine and the test pipeline is same.

from rcan.

arun121cs commented on July 28, 2024

Thanks a lot. You have helped me a big time. I was having an idea to use this model for a different type of data in an NLP related problem. Can you one very last time and i promise you for the last time tell me if i want to do that, what will i need to change in code and what form(like matrix or vector) i will need to prepare my dataset. Consider i have various texts and i converted them into numbers by some method.

…

On Tue, Jul 31, 2018 at 10:01 AM Yulun Zhang ***@***.***> wrote: Hi, The testing pipeline for the two degradation models keeps the same. I added '--degradation BD' for BD degradation for easier file naming. For example, an LR image is named as 'bird_LRBD_x2.png'. After super-resoltuion, we name the output as 'bird_RCAN_x2.png'. [image: image] <https://user-images.githubusercontent.com/22829440/43437825-d7865a98-943f-11e8-90d2-ab2f5c217ee2.png> If you don't use it, it's fine and the test pipeline is same. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Ahr4U1Fi026duDT6X8vIjM6fL-TalMLTks5uL92hgaJpZM4Vfh9M> .

-- Thanks and Regards, Arun Sharma 16MCMI04 M.Tech AI

from rcan.

arun121cs commented on July 28, 2024

I am training model with new images. I am converting images from HR to LR. Do i need to convert them into binary for that too?

…

On Tue, Jul 31, 2018 at 11:18 AM Arun Sharma ***@***.***> wrote: Thanks a lot. You have helped me a big time. I was having an idea to use this model for a different type of data in an NLP related problem. Can you one very last time and i promise you for the last time tell me if i want to do that, what will i need to change in code and what form(like matrix or vector) i will need to prepare my dataset. Consider i have various texts and i converted them into numbers by some method. On Tue, Jul 31, 2018 at 10:01 AM Yulun Zhang ***@***.***> wrote: > Hi, > The testing pipeline for the two degradation models keeps the same. > I added '--degradation BD' for BD degradation for easier file naming. For > example, an LR image is named as 'bird_LRBD_x2.png'. After > super-resoltuion, we name the output as 'bird_RCAN_x2.png'. > [image: image] > <https://user-images.githubusercontent.com/22829440/43437825-d7865a98-943f-11e8-90d2-ab2f5c217ee2.png> > > If you don't use it, it's fine and the test pipeline is same. > > — > You are receiving this because you modified the open/close state. > Reply to this email directly, view it on GitHub > <#5 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/Ahr4U1Fi026duDT6X8vIjM6fL-TalMLTks5uL92hgaJpZM4Vfh9M> > . > -- Thanks and Regards, Arun Sharma 16MCMI04 M.Tech AI

-- Thanks and Regards, Arun Sharma 16MCMI04 M.Tech AI

from rcan.

arun121cs commented on July 28, 2024

Processing Test: error: invalid concatenation of structure with matrix error: called from Prepare_TestData_HR_LR at line 18 column 19 I am getting this error.

…

On Thu, Aug 2, 2018 at 2:34 PM Arun Sharma ***@***.***> wrote: I am training model with new images. I am converting images from HR to LR. Do i need to convert them into binary for that too? On Tue, Jul 31, 2018 at 11:18 AM Arun Sharma ***@***.***> wrote: > Thanks a lot. You have helped me a big time. I was having an idea to use > this model for a different type of data in an NLP related problem. Can you > one very last time and i promise you for the last time tell me if i want to > do that, what will i need to change in code and what form(like matrix or > vector) i will need to prepare my dataset. Consider i have various texts > and i converted them into numbers by some method. > > On Tue, Jul 31, 2018 at 10:01 AM Yulun Zhang ***@***.***> > wrote: > >> Hi, >> The testing pipeline for the two degradation models keeps the same. >> I added '--degradation BD' for BD degradation for easier file naming. >> For example, an LR image is named as 'bird_LRBD_x2.png'. After >> super-resoltuion, we name the output as 'bird_RCAN_x2.png'. >> [image: image] >> <https://user-images.githubusercontent.com/22829440/43437825-d7865a98-943f-11e8-90d2-ab2f5c217ee2.png> >> >> If you don't use it, it's fine and the test pipeline is same. >> >> — >> You are receiving this because you modified the open/close state. >> Reply to this email directly, view it on GitHub >> <#5 (comment)>, >> or mute the thread >> <https://github.com/notifications/unsubscribe-auth/Ahr4U1Fi026duDT6X8vIjM6fL-TalMLTks5uL92hgaJpZM4Vfh9M> >> . >> > > > -- > Thanks and Regards, > Arun Sharma > 16MCMI04 > M.Tech AI > > -- Thanks and Regards, Arun Sharma 16MCMI04 M.Tech AI

-- Thanks and Regards, Arun Sharma 16MCMI04 M.Tech AI

from rcan.

Out of memory error about rcan HOT 12 CLOSED

Comments (12)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent