DoveG

Results 12 comments of DoveG

One possible solution to the OutOfMemoryError is to edit the parameters of "bs". It's in your startup shell: `--bs 8 --eval_bs 4` bs means "batch size", replacing it with a...

Based on the information you provided, the path is incorrect, maybe you can change the path in your *.sh file like this: `--eval_le models/rationale/predictions_ans_eval.json` However, the **real path** depends on...

> > Based on the information you provided, the path is incorrect, maybe you can change the path in your *.sh file like this: > > `--eval_le models/rationale/predictions_ans_eval.json` > >...

1. How to start the application on windows. The repository only provides bash to start the application, so I write bat to do that and it seems to work. This's...

2. CUDA out of memory. This is a very resource-intensive task, and my GPU doesn't seem capable of doing it. However, when I modify the parameter "bs" to a smaller...

3. Server jammed The model run out of the RAM of PC, and it get stuck. I have experienced long waits when the computer was on the verge of crashing....

> > 3. CUDA out of memory. > > Same problem. @bourneo could you provide your gpu memory spec? 6G VRAM. It may be helpful to set the batch size...

> You can also set the parameter eval_acc to a lower value to dump the accumulated results from the GPU to the RAM more often. Obviously, this needs a lot...

> > > You can also set the parameter eval_acc to a lower value to dump the accumulated results from the GPU to the RAM more often. Obviously, this needs...

Maybe you should set an environment variable "ARGOS_DEVICE_TYPE" in your operation system. Take windows for example: ![2023-03-31](https://user-images.githubusercontent.com/25973794/229034354-84169408-a5a7-4459-b1a8-55601ee3f9d7.jpg) More details are in this issue: #210 By the way, cuda must be...