search

paulm12 / forwardforward_python

Python translation of Hinton's "ForwardForward" algorithm
BSD 3-Clause "New" or "Revised" License
2 stars 0 forks source link

Directly load the MNIST dataset? #2

Open ajinkya-kulkarni opened 1 year ago

ajinkya-kulkarni commented 1 year ago

Hi, thanks for this repo! I had a short question - is possible to directly load the MNIST dataset using tensorflow as (train_images, train_labels), (test_images, test_labels) = tensorflow.keras.datasets.cifar10.load_data() and then run the notebook? Would be helpful, instead of loading the data from the .mat file.

ajinkya-kulkarni commented 1 year ago

The reason I am asking this is because I do not understand what are batchsize, batchtargets, finaltestbatchdata, finaltestbatchtargets etc. Therefore it would be helpful if the data is loaded from a standard library.

paulm12 commented 1 year ago

Hey Ajinkya, I will have these using standard libraries in PyTorch ASAP, probably within the end of the week. This version of the code only uses Hinton's MNIST data but the other versions I am working on will have autograd and be easy to swap out datasets (for an upcoming paper). -Paul

On Thu, Mar 30, 2023 at 4:40 AM Ajinkya Kulkarni @.***> wrote:

The reason I am asking this is because I do not understand what are batchsize, batchtargets, finaltestbatchdata, finaltestbatchtargets etc. Therefore it would be helpful if the data is loaded from a standard library.

— Reply to this email directly, view it on GitHub https://github.com/paulm12/forwardforward_python/issues/2#issuecomment-1490002564, or unsubscribe https://github.com/notifications/unsubscribe-auth/AF7RKPPU3W4C2CEVGEP44D3W6VIJZANCNFSM6AAAAAAWM72Z6M . You are receiving this because you are subscribed to this thread.Message ID: @.***>

ajinkya-kulkarni commented 1 year ago

Thank for the info! Can you update/comment here when you have it ready? I will have a look then - really interested to see it in action!