bug : Update gaussian_model.py 154-155 lines

[learnLearnDayUp]

Description : During code study, I found a minor issue in the create_from_pcd function when initializing SH features
Current code:

features = torch.zeros((fused_color.shape[0], 3, (self.max_sh_degree + 1) ** 2)).float().cuda()
features[:, :3, 0] = fused_color
features[:, 3:, 1:] = 0.0

Problem：
The second dimension of features is exactly 3 (RGB channels), so :3 is redundant
features[:, 3:, 1:] will produce an empty slice since dimension 1 only has indices 0,1,2

Proposed fix：

features = torch.zeros((fused_color.shape[0], 3, (self.max_sh_degree + 1) ** 2)).float().cuda()
features[:, :, 0] = fused_color
features[:, :, 1:] = 0.0

This sets the DC component (index 0) for all RGB channels to the initial color, and zeros out all higher-order SH coefficients. The functionality remains identical, but the code is cleaner and avoids unnecessary slicing.

Impact: No functional change, just code clarity improvement。