pytorch-faster-rcnn icon indicating copy to clipboard operation
pytorch-faster-rcnn copied to clipboard

How to run the code on pytorch1.7 Nvidia 3090?

Open Hbenbenyu opened this issue 4 years ago • 4 comments

Hi,it is very great of you to do the contribution,but I want to run the code on the Nvidia 3090,how to modify the code?Thanks too much if you can help me.

Hbenbenyu avatar Nov 12 '20 14:11 Hbenbenyu

Hi,I use pytorch1.7 and train on 2080. I think this code can be directirely run on the 3090, you can have a try. If there is any mistake, please let me know first, you need to download coco dataset

AlphaJia avatar Nov 14 '20 14:11 AlphaJia

Fine,thank you.I will try to use coco dataset. But my dataset format is voc,and I need to use the pre-trained model. Anyway, it is great of you to answer my question patiently.

Hbenbenyu avatar Nov 16 '20 03:11 Hbenbenyu

Hi,I download the COCO dataset and set the parameters,then I run the code.But there are some questions. 1.what's the meaning of the parameter 'resume = '' 'in train_config.py? 2.the Epoch is not the parameter which I set in train_configs.py when run the code, it's always 19715; 3.the code stops when it runs to the '------>Starting training data valid', so the weigths can't be saved . Thank you very much if you can answer my question. I also ask such questions in the 知乎 ,maybe we can chat there.=-=

Hbenbenyu avatar Nov 18 '20 14:11 Hbenbenyu

Hi,I download the COCO dataset and set the parameters,then I run the code.But there are some questions. 1.what's the meaning of the parameter 'resume = '' 'in train_config.py? 2.the Epoch is not the parameter which I set in train_configs.py when run the code, it's always 19715; 3.the code stops when it runs to the '------>Starting training data valid', so the weigths can't be saved . Thank you very much if you can answer my question. I also ask such questions in the 知乎 ,maybe we can chat there.=-=

I met the same problem, did you sovle it?

shuxiao0312 avatar Jan 14 '21 08:01 shuxiao0312