forked from ccf-ai-infra/GPUCodeForces
63 lines
2.0 KiB
Python
63 lines
2.0 KiB
Python
import torch
|
|
import torch.nn as nn
|
|
|
|
class Model(nn.Module):
|
|
"""
|
|
Minkowski Distance implementation.
|
|
Computes the Minkowski distance between two sets of vectors with order p.
|
|
"""
|
|
def __init__(self, p=2):
|
|
super(Model, self).__init__()
|
|
self.p = p
|
|
if p <= 0:
|
|
raise ValueError("p must be positive")
|
|
|
|
def forward(self, x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
|
|
"""
|
|
Compute Minkowski distance between x and y.
|
|
|
|
Args:
|
|
x (torch.Tensor): First set of vectors [batch_size, feature_dim]
|
|
y (torch.Tensor): Second set of vectors [batch_size, feature_dim]
|
|
|
|
Returns:
|
|
torch.Tensor: Minkowski distances [batch_size]
|
|
"""
|
|
# Input validation
|
|
if x.shape != y.shape:
|
|
raise ValueError(f"Input tensors must have the same shape, got {x.shape} and {y.shape}")
|
|
|
|
if x.dim() != 2:
|
|
raise ValueError(f"Input tensors must be 2D, got {x.dim()}D")
|
|
|
|
# Compute absolute differences
|
|
abs_diff = torch.abs(x - y)
|
|
|
|
# Compute Minkowski distance: (Σ|x_i - y_i|^p)^(1/p)
|
|
if self.p == 1:
|
|
# Manhattan distance
|
|
minkowski_dist = torch.sum(abs_diff, dim=1)
|
|
elif self.p == 2:
|
|
# Euclidean distance
|
|
minkowski_dist = torch.sqrt(torch.sum(abs_diff ** 2, dim=1))
|
|
elif self.p == float('inf'):
|
|
# Chebyshev distance
|
|
minkowski_dist = torch.max(abs_diff, dim=1)[0]
|
|
else:
|
|
# General Minkowski distance
|
|
minkowski_dist = torch.pow(torch.sum(torch.pow(abs_diff, self.p), dim=1), 1.0/self.p)
|
|
|
|
return minkowski_dist
|
|
|
|
batch_size = 256
|
|
feature_dim = 512
|
|
|
|
def get_inputs():
|
|
# Generate two sets of vectors
|
|
x = torch.randn(batch_size, feature_dim)
|
|
y = torch.randn(batch_size, feature_dim)
|
|
return [x, y]
|
|
|
|
def get_init_inputs():
|
|
return [2] # p value (default: Euclidean distance)
|