Need help - Takes forever on newly setup machine

[asawant (Anup)]

Hi,

I just setup a new machine of my own with core i5 and gtx 1070. I installed,NVIDIA drivers, CUDA and did fastai setup. However, when I am trying to run the example CATS and Dogs model, its taking forever to return any results after downloading the resnet model. How do I make sure the learner is using GPU and not CPU ?

[thiago]

You can run in your terminal: watch -n 1 nvidia-smi

Check out the Volatile GPU-Util column.

[asawant (Anup)]

Thanks !! It says 0% Default but I do see it as one of the GPU processes at the bottom. Am I missing something ? Why is it taking so much time ?

[thiago]

Cool, now you could test if pytorch is using your GPU.

Check it out:

How to check your pytorch / keras is using the GPU?

As we work on setting up our environments, I found this quite useful: To check that torch is using a GPU: In [1]: import torch In [2]: torch.cuda.current_device() Out[2]: 0 In [3]: torch.cuda.device(0) Out[3]: <torch.cuda.device at 0x7f2132913c50> In [4]: torch.cuda.device_count() Out[4]: 1 In [5]: torch.cuda.get_device_name(0) Out[5]: 'Tesla K80' To check that keras is using a GPU: import tensorflow as tf tf.Session(config=tf.ConfigProto(log_device_placement=True)) and check the jupyte…

[asawant (Anup)]

Yes, it does look like pytorch is using GPU from the steps mentioned in that thread.

[gerardo (Gerardo Garcia)]

How long is a lot?
Loading the images take time
5 - 7 min aceptable
More maybe not

[asawant (Anup)]

Its working now. It turned out, just adding /usr/local/cuda-9.0 to PATH didn’t help. I now added a variable CUDA_HOME to .bashrc. Thanks for the pointers !!