Fix off by one error in inference.py

[gabrieltseng]

Currently, the inference will miss a prediction (and fail during dataframe creation) if x_np.shape[0] % batch_size == 1.

This PR fixes that. Specifically:

>>> import numpy as np
>>> total_size, batch_size = 1185857, 64
>>> x_np = np.ones(total_size)
>>> # current approach
>>> batches = [x_np[i : i + batch_size] for i in range(0, x_np.shape[0] - 1, batch_size)]
>>> sum([len(x) for x in batches])
1185856
>>> # fixed
>>> batches = [x_np[i : i + batch_size] for i in range(0, x_np.shape[0], batch_size)]
>>> sum([len(x) for x in batches])
1185857

[ivanzvonkov]

Nice catch, thank you!