facok
0973fd0fd8
Coverage: 29% → 100% (41/41 public items documented). Keeps the existing concise style with inline shape annotations.
66 lines
3.6 KiB
Python
66 lines
3.6 KiB
Python
"""Hardcoded alpha_t and beta_t tables from paper Appendix F (51 entries, t=0..50)."""
|
||
|
||
import torch
|
||
|
||
# Shift alpha_t: 3D vectors for each timestep t=0..50
|
||
ALPHA_T = [
|
||
[2.3413, -2.3586, 0.4266], [2.3574, -2.3833, 0.4644], [2.3638, -2.3904, 0.4883],
|
||
[2.3734, -2.3951, 0.5122], [2.3831, -2.3993, 0.5384], [2.3925, -2.4026, 0.5647],
|
||
[2.4023, -2.4047, 0.5919], [2.4124, -2.4060, 0.6198], [2.4226, -2.4064, 0.6484],
|
||
[2.4330, -2.4060, 0.6772], [2.4437, -2.4051, 0.7065], [2.4546, -2.4035, 0.7367],
|
||
[2.4659, -2.4011, 0.7668], [2.4775, -2.3981, 0.7974], [2.4897, -2.4009, 0.8312],
|
||
[2.5021, -2.4036, 0.8656], [2.5148, -2.4065, 0.9008], [2.5277, -2.4093, 0.9364],
|
||
[2.5408, -2.4123, 0.9727], [2.5542, -2.4154, 1.0099], [2.5680, -2.4186, 1.0481],
|
||
[2.5820, -2.4218, 1.0868], [2.5963, -2.4252, 1.1263], [2.6110, -2.4288, 1.1672],
|
||
[2.6261, -2.4324, 1.2090], [2.6416, -2.4363, 1.2520], [2.6575, -2.4403, 1.2957],
|
||
[2.6738, -2.4444, 1.3406], [2.6904, -2.4485, 1.3865], [2.7074, -2.4529, 1.4336],
|
||
[2.7250, -2.4574, 1.4818], [2.7432, -2.4621, 1.5314], [2.7618, -2.4669, 1.5823],
|
||
[2.7810, -2.4720, 1.6344], [2.8006, -2.4771, 1.6878], [2.8209, -2.4826, 1.7430],
|
||
[2.8418, -2.4883, 1.7995], [2.8631, -2.4944, 1.8578], [2.8853, -2.5005, 1.9179],
|
||
[2.9080, -2.5066, 1.9793], [2.9313, -2.5132, 2.0426], [2.9555, -2.5199, 2.1082],
|
||
[2.9804, -2.5268, 2.1756], [3.0060, -2.5338, 2.2450], [3.0328, -2.5411, 2.3172],
|
||
[3.0603, -2.5486, 2.3914], [3.0889, -2.5561, 2.4682], [3.1189, -2.5640, 2.5482],
|
||
[3.1497, -2.5725, 2.6302], [3.1824, -2.5796, 2.7175], [3.2152, -2.5889, 2.8050],
|
||
]
|
||
|
||
# Scale beta_t: 3D vectors for each timestep t=0..50
|
||
BETA_T = [
|
||
[0.0163, 0.0172, 0.0295], [0.0905, 0.0716, 0.0999], [0.1345, 0.1123, 0.1544],
|
||
[0.1826, 0.1491, 0.2065], [0.2360, 0.1899, 0.2630], [0.2904, 0.2316, 0.3202],
|
||
[0.3471, 0.2749, 0.3793], [0.4050, 0.3191, 0.4394], [0.4640, 0.3641, 0.5003],
|
||
[0.5231, 0.4091, 0.5611], [0.5834, 0.4547, 0.6228], [0.6456, 0.5016, 0.6861],
|
||
[0.7077, 0.5481, 0.7488], [0.7713, 0.5958, 0.8127], [0.8410, 0.6496, 0.8866],
|
||
[0.9119, 0.7044, 0.9616], [0.9845, 0.7605, 1.0386], [1.0578, 0.8172, 1.1163],
|
||
[1.1325, 0.8750, 1.1957], [1.2094, 0.9344, 1.2771], [1.2880, 0.9953, 1.3606],
|
||
[1.3680, 1.0571, 1.4453], [1.4498, 1.1205, 1.5321], [1.5341, 1.1858, 1.6216],
|
||
[1.6206, 1.2526, 1.7131], [1.7094, 1.3214, 1.8072], [1.7998, 1.3913, 1.9030],
|
||
[1.8927, 1.4633, 2.0014], [1.9879, 1.5370, 2.1022], [2.0854, 1.6126, 2.2056],
|
||
[2.1853, 1.6900, 2.3114], [2.2881, 1.7696, 2.4202], [2.3939, 1.8515, 2.5321],
|
||
[2.5021, 1.9354, 2.6467], [2.6133, 2.0215, 2.7642], [2.7280, 2.1106, 2.8857],
|
||
[2.8455, 2.2017, 3.0101], [2.9668, 2.2957, 3.1386], [3.0921, 2.3929, 3.2712],
|
||
[3.2204, 2.4922, 3.4067], [3.3523, 2.5946, 3.5464], [3.4888, 2.7006, 3.6911],
|
||
[3.6292, 2.8097, 3.8398], [3.7741, 2.9222, 3.9931], [3.9247, 3.0394, 4.1527],
|
||
[4.0793, 3.1597, 4.3168], [4.2393, 3.2843, 4.4866], [4.4053, 3.4142, 4.6636],
|
||
[4.5760, 3.5480, 4.8461], [4.7541, 3.6886, 5.0383], [4.9407, 3.8364, 5.2390],
|
||
]
|
||
|
||
# Pre-convert to tensors (lazily cached on first access)
|
||
_alpha_tensor = None
|
||
_beta_tensor = None
|
||
|
||
|
||
def get_alpha_table():
|
||
"""Return α_t table as tensor [51, 3], cached after first call."""
|
||
global _alpha_tensor
|
||
if _alpha_tensor is None:
|
||
_alpha_tensor = torch.tensor(ALPHA_T, dtype=torch.float32) # [51, 3]
|
||
return _alpha_tensor
|
||
|
||
|
||
def get_beta_table():
|
||
"""Return β_t table as tensor [51, 3], cached after first call."""
|
||
global _beta_tensor
|
||
if _beta_tensor is None:
|
||
_beta_tensor = torch.tensor(BETA_T, dtype=torch.float32) # [51, 3]
|
||
return _beta_tensor
|