[FEATURE] Add Multi Scale SSIM

[begumcig]

✨ What You’ll Do

Adding more metrics makes Pruna’s evaluations more insightful, so we’re excited to welcome Multi Scale SSIM from torchmetrics source into our metric lineup! 🎉

Multi Scale SSIM is a generalization of the SSIM score, which computes the SSIM score at different resolution scales to get a more accurate representation

---

🛠️ Dependency

We are already actively using metrics from torchmetrics so you should be good to go!

---

📐 Metric Class

We already have a wrapper class for integrating metrics from torchmetrics at pruna/evaluation/metrics/metric_torch.py. You can easily integrate MultiScale SSIM there!

Quick Overview

1. File & Class
   • Add MultiScale SSIM to the TorchMetrics Enum
2. Configuration
   • Choose a sensible call_type 
   • Decide if you need to implement an update function
3. Testing & Docs
   • Write unit tests under tests/evaluation/metrics/test_torch_metrics.py

And that's all!

✅ Acceptance Criteria

• Correct Output: compute(...) returns a sensible result
• Tests & Docs: All existing and new unit tests pass, and the documentation is updated

And don’t forget to give us a ⭐️!

---

❓ Questions?

Feel free to jump into our Discord channel if you hit any roadblocks. Can’t wait to see your contribution of this metric! 🚀

---

Share on X

[kris70lesgo]

@begumcig can u assign this issue to me ?

[minettekaum]

Hi @kris70lesgo 👋
If you’re participating in Hacktoberfest, just keep in mind that PRs are eligible starting tomorrow: https://hacktoberfest.com/participation/ :)

[kris70lesgo]

thanks for reminding me !

[R055A]

I have created a PR #361 for this issue. I would like to mention that there is no metrics module in tests/evaluation