Python `cbor` encode np.float32 as float

Question

Trying to encode in cbor some lists and dicts which may have deeply nested numpy float32 values.

Although those values looks similar and interchangeable:

import numpy as np
x = np.float32(1.001)
y = 2.002
print(x, y) # 1.001 2.001
print(type(x), type(y)) # <class 'numpy.float32'> <class 'float'>

cbor would fail to encode np.float32:

import cbor
cbor.dumps([x, y])
# ValueError: cannot serialize unknown object: 1.001

I don't want or need to retain the np.float32 type information, so I would be ok if cbor treats a np.float32 (and similar) object just like a basic float object.

But I don't want to deeply scan and convert my object (which in this simplified case is [x, y], but it is more complex in reality) as that would be a performance penalty.

Is there a way to register an encoder for np.float32 so that I can convert the value to Python's float and encode it with default cbor ancoder?