Disk or RAM

[Nadian-Ali]

Hi.. i am trying to run your code... i have 32 GB of RAM and more than 100GB of disk space. 3 GB of my RAM gets occupied for loading the dictionary and i get stuck there.... do i need a bigger RAM? Like 64GB of RAM space?

[LeeDoYup]

I remember
VQA implementations require 150 - 200 GB RAM (also in pytorch implementation)

It is because the implementations use pretrained features and load at once..
If you want to save, find or implement a data loader that satisfy

1. Load an image file at each time
2. Detect objects in image
3. and run VQA models.

The reason that many implementations use pretrained features (including this repo) is to save inference time during 1)-2) .

[LeeDoYup]

https://github.com/LeeDoYup/bottom-up-attention-tf/blob/master/README.md#prerequisites

[Nadian-Ali]

Thank you very much... i am trying to use virtual RAM... I hope it works

…

On Tue 3 Mar 2020, 4:19 PM Doyup Lee ***@***.*** wrote: Closed #2 <#2>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AN5NVTLAF4EOP7WC7WPTJ3LRFT4GNA5CNFSM4LABSJS2YY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOXBGGT6Q#event-3092015610>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AN5NVTMJVH2BWRJHJVAOIBTRFT4GNANCNFSM4LABSJSQ> .

[LeeDoYup]

@Nadian-Ali Okay. FYI, i achieve the VQA accuracy similar to pytorch implementation.
I implemented this model for my research, but recently i start to use Pytorch.
So i would not update this implementation (only maintenance).

Also, we can refer FAIR's awesome open source library for vqa: Pythia

[Nadian-Ali]

Thank you for your helpful information. I hope i dont have to switch to pytorch

…

On Fri 6 Mar 2020, 11:51 AM Doyup Lee ***@***.*** wrote: @Nadian-Ali <https://github.com/Nadian-Ali> Okay. FYI, i achieve the VQA accuracy similar to pytorch implementation. I implemented this model for my research, but recently i start to use Pytorch. So i would not update this implementation (only maintenance). Also, we can refer FAIR's awesome open source library for vqa: Pythia <https://learnpythia.readthedocs.io/en/latest/tutorials/quickstart.html> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AN5NVTPAD6S3M7ZQF43JHYTRGCW7TA5CNFSM4LABSJS2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEOAP27A#issuecomment-595656060>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AN5NVTIUOO2NKLPMYWHUXS3RGCW7TANCNFSM4LABSJSQ> .