Hi All,
I have 2 questions:
(1) About creating the dataset. after we convert all json files into a list (say called lst, s.t: lst[:][0} is a graph type and lst{:}{1} is the int ) , do i still need to define it sonehow as a dataset object or i can just use it as the input list for the dataloader . and if so, how do i do it? i think in the tutorial we didn't use existing files but rather created random graphs as the dataset...
(2) when creating the Databunch with fastai (like in the tutorial), i must have a valid_dl input. but in the Homework we just have test and train gaphs.
Thanks
Ori
Hi Ori,
I recommend that you create a dataset object that takes the list you created and returns the graph and label in the x,y format. you can technically use a list in the dataloader (without creating a dataset object), but each item in the list must be of the format (x,y) and the list can't be 2 dimensional list what you describe.
when you create the databunch just provide the test dataloader as a valid dataloader.