New Django field to store numpy ndarray.
Store a numpy n-dimensional array in database (compatible with all database backend). Use numpy save/load, you can define a shape (not required), and a dtype (default float32).
import numpy as np
from django.db import models
from ndarraydjango.fields import NDArrayField
class MyModel(models.Model):
vec1 = NDArrayField(shape=(32, 4), dtype=np.float64)
date = models.DateTimeField(auto_now_add=True)- dtype: the ndarray dtype (default np.float32)
- shape: the ndarray shape (default None)
- binary_serialize: when dump data, serialize to binary (base64) or json lists (default False)
This field type does not replace a static file storage. The main goal is to store parameter data, results of algorithms and small and medium machine learning models. A good indication is the shape of the nd-array. It would be static, and with a reasonable size. The overrall data size should not exceed 1mb. For example a field of 300x400 of 2 float32 value ( (300, 400, 2) dtype=float32) should be a maximum.