fix: correct variable name in pixel_tokens_with_pos_embed reshape operation

[lixiaolei1982]

Description

Fixed incorrect variable name in pixel_tokens_with_pos_embed reshaping operation where num_h_patches_per_window was used instead of num_w_patches_per_window for the width dimension.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Changes Made

• Change: num_h_patches_per_window → num_w_patches_per_window in the fourth dimension of reshape operation
• File: rfdetr/models/backbone/dinov2_with_windowed_attn.py
• Line: ~315

Code Comparison

Before:

windowed_pixel_tokens = pixel_tokens_with_pos_embed.reshape(
    batch_size * num_windows, num_h_patches_per_window, 
    num_windows, num_h_patches_per_window, -1)

**After:**
```python
windowed_pixel_tokens = pixel_tokens_with_pos_embed.reshape(
    batch_size * num_windows, num_h_patches_per_window, 
    num_windows, num_w_patches_per_window, -1)

## Impact
- Improves code readability and consistency
- No functional changes

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

This PR fixes a bug in the tensor reshape operation where the incorrect variable was used for the width dimension when partitioning image patches into windows.

Changes:

• Fixed variable name from num_h_patches_per_window to num_w_patches_per_window in the 4th dimension of the reshape operation

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.