search

airsplay / lxmert

PyTorch code for EMNLP 2019 paper "LXMERT: Learning Cross-Modality Encoder Representations from Transformers".
MIT License
923 stars 157 forks source link

training killed while loading image features #98

Open Youngju-Na opened 3 years ago

Youngju-Na commented 3 years ago

I've been following the vqa fine tuning process but I constantly find problems here. I successfully finished the bash run/vqa_finetune.bash 0 vqa_lxr955_tiny --tiny command.

but I found this issue.

image

I used wsl2 (ubuntu 18.04 tsl) with rtx3080

Thanks for your comments.

yezhengli-Mr9 commented 3 years ago

Hardware might not be enough. I remember --tiny is not hard to run (for example, macbook with 16 GB CPU-only), it is runnable but just much slower than GPU-based training.

yezhengli-Mr9 commented 3 years ago

Hardware might not be enough. I remember --tiny is not hard to run (for example, macbook with 16 GB CPU-only), it is runnable but just much slower than GPU-based training.

Ok, I do not know since you have GPU rtx3080.

LeungWaiHo commented 2 years ago

I've been following the vqa fine tuning process but I constantly find problems here. I successfully finished the bash run/vqa_finetune.bash 0 vqa_lxr955_tiny --tiny command.

but I found this issue.

image

I used wsl2 (ubuntu 18.04 tsl) with rtx3080

Thanks for your comments.

yes! I use rtx2080ti, having the same question.

LeungWaiHo commented 2 years ago

Hardware might not be enough. I remember --tiny is not hard to run (for example, macbook with 16 GB CPU-only), it is runnable but just much slower than GPU-based training.

Ok, I do not know since you have GPU rtx3080.

Could I ask which GPU or how many GPUs do you use when vqa fune tuning? thanks!