Ones — Constant Ones Weight Initializer¶
Fills a weight tensor with constant ones.
Overview¶
Ones returns an array of the requested shape filled entirely with 1.0. It is most commonly used to initialize bias-like terms or gating variables that should start at a non-zero unit value (for example, the scale parameter s in gnostic neurons).
Key Features¶
- Constant value of
1.0across the entire tensor - Useful for bias/gate/scale initialization
- Shape validated against the initializer helpers
- Returns
np.float64array by default
Parameters¶
| Parameter | Type | Default | Description |
|---|---|---|---|
dtype |
np.dtype |
np.float64 |
Output data type |
Methods¶
__call__(shape)¶
Return an array of ones with the requested shape and validated dimensions.
Example Usage¶
from machinegnostics.magnet import Ones
init = Ones()
b = init((20,))
print(b.shape) # (20,)
print(b[:3]) # [1. 1. 1.]
Author: Nirmal Parmar