Is it normal to be so slow in train phase？ about crslab HOT 4 CLOSED

Could you please provide the full log for this command? The data process step just cost 10 seconds on my machine.

from crslab.

Could you please provide the full log for this command? The data process step just cost 10 seconds on my machine.

Log

cpu usage, this process exhaust all my cpus.

My cpu
$ cat /proc/cpuinfo | grep name | cut -f2 -d: | uniq -c 32 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz

from crslab.

Could you please provide the full log for this command? The data process step just cost 10 seconds on my machine.

Log

cpu usage, this process exhaust all my cpus.

My cpu
$ cat /proc/cpuinfo | grep name | cut -f2 -d: | uniq -c 32 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz

The log shows that it cost almost 7 hours in the training stage. I guess that you're not using the GPU to run this process.

You can:

1. Run command "nvidia-smi" to ensure the process 29050 are running on the GPU;
2. If there isn't any process running on the GPU, you can run the following command to verify if the cuda toolkit is available.

python -c "import torch; print(torch.cuda.is_available())"

from crslab.

Could you please provide the full log for this command? The data process step just cost 10 seconds on my machine.

Log

cpu usage, this process exhaust all my cpus.

My cpu
$ cat /proc/cpuinfo | grep name | cut -f2 -d: | uniq -c 32 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz

The log shows that it cost almost 7 hours in the training stage. I guess that you're not using the GPU to run this process.

You can:

1. Run command "nvidia-smi" to ensure the process 29050 are running on the GPU;
2. If there isn't any process running on the GPU, you can run the following command to verify if the cuda toolkit is available.

python -c "import torch; print(torch.cuda.is_available())"

you are right!!!!!!
I has thought data process has no need of GPU, but it need.
Thank you

from crslab.