Update dataloaders.py remove float64 shapes

[glenn-jocher]

May help #8862

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced data loader efficiency by simplifying shape arrays.

📊 Key Changes

• Removed explicit dtype=np.float64 specification from np.array(shapes) assignment.

🎯 Purpose & Impact

• 🚀 Purpose: The change is intended to simplify the data loading process by not explicitly defining the data type of the shape array. This could potentially lead to using a default data type that's more memory-efficient.
• ✅ Impact: Users might experience improved memory usage and possibly faster data loading times when training their models, especially if the default data type is smaller than float64.