Sample data first before load to numpy array

[BirkhoffG]

Before, even if sample_frac is set, we use the entire data for building train_dataset, val_dataset, test_dataset.

When data is very large, this approach could unwantedly blow the memory because building datasets ( train_dataset etc.) assign new memory.

Now, I only load data with first sample_frac of data.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB