Installing path in Dl SDK

Installing path in Dl SDK

Hi,

Sorry for such a naive question, but how can I set an installing path for DL SDK? When I first installed it, dl-genie appeared in the home directory and even then in order to change anything in it I had to use sudo. I want to be able to install it in an arbitrary path. I looked into Intel_Deep_Learing_SDK_Training_Tool_Installation_Guide but didn't find anything related.

 

Thank you

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

I was wondering why you need to go in such lower levels. DLSDK training tool comes with the web UI to abstract low level details and let you focus on the deep learning aspects. All our documentation are focused on explaining the web UI and how you can use it to train deep learning models.

Hello,

"~/dl-genie" is a path which is mounted as a volume inside the docker container of the training tool. You can re-run our installer and change this path, see "Mount file system path" option in the UI installer and "-volume" option in the command line install script for Linux.

Please let me know if it doesn't work for you.

Thanks,
Alexey

@Pubudu S.

The reason is quite simple! I don't want an extra folder in my home directory :)

@Alexey S.

I did what you said and it's installed in the path I wanted, but I've got some issues now!

When I try to fine tune a googlenet I can't see an option to specify where the pretrained weights are! and so running results in error (saying File not found: /opt/caffe/models/pretrained/bvlc_googlenet.caffemodel, though I cannot find such an address in the first place).

Another thing that is rather annoying is that when I click duplicate to copy the model I have to tune nearly all the parameters (like learning rate).

And one other question! Isn't there a way to test a previously trained model with this tool? rather than training a whole new one?

Thank you!

Leave a Comment

Please sign in to add a comment. Not a member? Join today