[Help] Alternatives to AWS and Microsoft Azure

Well, no I haven’t gotten far, I just got the environment setup yesterday the way I wanted it. But, I don’t expect any problems now.

If you run into any other stuff, just lmk, I should be here all day.

Ok. All is working with keras now installed.


1.Any idea what this warning is about and whether I should be concerend?

DEBUG: nvcc STDOUT nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be remove
d in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
Creating library C:/Users/<username>/AppData/Local/Theano/compiledir_Windows-10-10.0.14393-SP0-Intel64_Family_6_Model_9
4_Stepping_3_GenuineIntel-3.5.2-64/tmps2ta3n9j/m91973e5c136ea49268a916ff971b7377.lib and object C:/Users/<username>/AppDat

2.What are these warnings and is there a way to suppress them?

> DEBUG: nvcc STDOUT mod.cu
> Creating library C:/Users/<username>/AppData/Local/Theano/compiledir_Windows-10-10.0.14393-SP0-Intel64_Family_6_Model_9
> 4_Stepping_3_GenuineIntel-3.5.2-64/tmp6z22v7rz/m4b03c91158bcf78ba154a9d2acbfd1bc.lib and object C:/Users/<username>/AppDat
> a/Local/Theano/compiledir_Windows-10-10.0.14393-SP0-Intel64_Family_6_Model_94_Stepping_3_GenuineIntel-3.5.2-64/tmp6z22v7
> rz/m4b03c91158bcf78ba154a9d2acbfd1bc.exp

The first time I run python.\minst_cnn.py’ I get a million of these debug statements. I don’t know why and/or whether I should be concerned.

I didn’t install OpenBLAS and running against the 960M I see the follow time/epoch:

Using CPU: Forget it about (i.e., forever)
Using GPU: 25s
Using GPU + cuDNN: 9s

These aren’t as good as you are getting with graphics card but they seem pretty decent for the 960M running on an XPS laptop, no?

The author did mention one of the warnings which was coming from using a cuDNN that is above the recommended for Theano, however, the speed improved and accuracy stayed constant, so, seems all good.

As for the others, I have no idea whats going with the deprecated arch’s but it does say you can suppress the warning with a flag. I only got that I think on the first run. Sorry that it took me awhile to respond, I was newUser limited. :stuck_out_tongue:

As for your results they look pretty damn good for a 960m. better than I would’ve guessed.

1 Like

@DBAggie https://education.github.com/pack

I think this might be useful for you to be able to afford AWS for the purposes of this class. It comes with ~$150 of free AWS time.

Can you please provide the link for setting up without a GPU. Not finding that anywhere.

Cheers… Ange

@no_angel To set up without using a GPU, follow the AWS setup video, only use the setup_t2.sh script instead of the p2 script.

@wgpubs Don’t worry about the warnings. If you do want to suppress them, I found this approach although I haven’t tried it myself

I actually messaged the theano people … and the issue is resolved in the dev branch. Verified that all works against it thus far.

1 Like

Ok people, this is container working on cloud9, works for deeplearning course udacity, szhould work here. time will be longer as this is only cpu, but works without problems just longer, I have used it to solve easy kaggle competitions with it https://community.c9.io/t/tensorflow-deep-learning-artificial-intelligence-ai-neural-networks/5203 enjoy

1 Like

Just to make sure I understand it all correctly and please bear with me if it was clearly stated already.

  1. I can have my own system (NOT AWS) and be able to fully follow this course even though my system doesn’t have an NVIDIA GPU, but only an Intel integrated GPU.
  2. I will be able to do all the tasks, but with the CPU, which will be much slower but still doable.

Are these assumptions are correct?


sorry, thwer is only tensorflow backend so no use for this course

I came across this today and thought it looked interesting (100 free hours!)


getting the data set downloaded seems like a hassle however.

Just use google compute. Its economical

I wanted to follow along the course really bad and was in a similar situation to you. After a week or so of wrangling with nvidia drivers and a few other issues, I finally installed it. I made a setup guide so others wouldn’t have to go through the same tortuous experience. Feel free to go through it, fork it and add your own issues. Most of the problems I faced were to do with installing the nvidia drivers on my laptop, the rest was relatively a breeze.

1 Like