I created a MNIST boiler plate ipython notebook and wanted to share - maybe it will be helpful to someone. It can be of use in following scenarios:
- when you only have 20 - 30 minutes which might be too little to do anything useful on AWS with the cats & dogs dataset
- when you are on the road and don’t have stable Internet connection
- when you are struggling to make sense of lesson 1 and 2 - this can allow you to mess with things foregoing for the time being the complexity of loading the data, connecting to aws, etc
- if you want to experiment with this rather cool dataset / have fun
I am myself planning to use it to continue experimenting with keras. I have a hard time finding blocks of time in my day where I can devote 1 - 2 hrs straight to working on the course and this is probably the amount of time required when one wants to work with the larger datasets on AWS. Also, this gives me a really quick feedback loop for experimenting which is quite nice given that I am a total newb to keras - this experience should make it easier for me to work on the bigger models on AWS.
If you would be interested in giving this a shot, running the first cell will load things up for you and all you need to do to get going is add at least a single layer in cell nr 2 where it says TODO. Once you have this, you can run all the remaining cells.