As anyone managed to test the packages from this distribution, such as numpy and/or scipy, in AWS Lambda? If yes, would be very interested in understanding the steps needed to achieve this.
@Intel, would be great for users to have a github repository, as https://github.com/vitolimandibhrata/aws-lambda-numpy, for the optimized packages.
We have not looked at Lambda, but it looks like a good idea. We will see if there is a way to install a MKL-optimized numpy. I am not sure how long it will take so if someone else has success, please share it. Thanks.
Thanks for taking a look. I guess with Numpy and Pandas many developers will be happy.
If I could add one more thing to the wish list is a docker image that includes the (optimized) intel python distribution.
Nonetheless, I am positively surprised with the performance of the distribution. Well done!
It took me some days, but I managed to make it work in AWS Lambda. Using an Amazon Linux AMI, I set up a virtual environment pointing to the Intel Python 2.7 interpreter. Then I simply used pip to install Pandas and Numpy (got an error first as I didnt have some developer tools).
Then I just zipped the virtual environment folder, and now I can import Numpy and Pandas in AWS Lambda. The total size of the zip file is around 25MB, so overall it is not a problem. There is probably a better way, ie, delete the test cases, etc, but at least it got me going with the library.
I also tested f2py and it also worked like a charm in AWS Lambda. If there are some benchmarks that you would like me to run, just let me know. Happy to help.
Will try to replicate the steps for the 2017 beta distribution in the next couple of days, and will post each and every step.