Problem: "Out of Memory" when HOReIDTrainer.py runs

nickhuang1996 / HJL-re-id

A awesome project for person re-id. There are re-implements of PCB, MGN, MDRS(ours), PGFA, Pyramidal and HOReID.

MIT License

53 stars 17 forks source link

Problem: "Out of Memory" when HOReIDTrainer.py runs #4

Open qianqiansbaby opened 4 years ago

qianqiansbaby commented 4 years ago

Sorry to bother you because I have some questions on training for HOReID. When I run the code for HOReID and training in some epochs then it occurs errors: Out of Memory. I check that the problem is not raised from CUDA out of memory. And the batch_size is set to 4 but no use. Please tell me why your code has the memory errors.

nickhuang1996 commented 4 years ago

@qianqiansbaby Please check if :point_right:RAM:point_left: or :point_right:virtual memory:point_left: of your computer is enough. You are supposed to set virtual memory from 16GB to 32GB if you do not need to but a larger RAM.

ssbilakeri commented 4 years ago

@qianqiansbaby Hi kill some processes by pressing alt+ctr+del and you can also reduce batch size This method worked for me

D77-d77 commented 4 years ago

@qianqiansbaby PGFATrainer works well! You can learn this models for occluded person re-id tasks. HOReID needs much RAM and GPU resources.