[FEATURE] Loading a Dataset as a tuple instead of a dictionary

🚨🚨 Feature Request

• A new implementation (Improvement, Extension)

Is your feature request related to a problem?

Currently, the generators for the Keras library uses batches of (x, y) tuples in order to feed data to the model. The Hub schemas only let us change the shape and/or datatype of the data within a dictionary. If we could extend this behavior to load data as batches of tuples as well the keras integration would be seamless.

If your feature will improve HUB

Hub would be seamlessly integrated with Keras. End-Users would not need to modfiy the data further after loading it via hub, when they are using Keras models and generators.

Description of the possible solution

We might need to restructure the APIs that are providing the data. Though unlikely, could be breaking.

[mynameisvinn]

Gotcha, so something more like this?

def load_data(ds, batch_size):
"""Return a generator to yield (X, y).
"""
   for i in range(n, batch_size):
       X = ds["data", i: i + batch_size].compute()
       y = ds["label", i: i + batch_size].compute()
       yield X, y

[DebadityaPal]

Thanks, @mynameisvinn this works perfectly!

@AbhinavTuli what I was initially doing was something like this, I converted the whole data to (X, y) pairs first.

def convert(...):
     ``` converts the whole data into (X,y)```
     ...
     return (X, y)

(X, y) =  convert(ds)
model.fit(X, y)

what I would like to do is:

```suppose this sets the schema to return tuples```
schema = (np.ndarray(int64), np.ndarray(int64))   

ds = Dataset("activeloop/mnist", mode='w', schema=schema)
 
model.fit(ds, batch_size=64)

I know that schema is supposed to give a structure to a local dataset that one is about to upload, but it would be good if we could use it to restructure retrieved datasets and especially convert them to tuples from dictionaries. However, this problem is completely solved by the code snippet provided by @mynameisvinn, so I don’t think it would be necessary to change the current flow of the software.