🐛 Fix issue with MPS device

[Jiaqi-Lv]

This PR fixes #1070.
It standardizes the way tensors are moved to a device and cast to float32 across multiple model architectures in the codebase. The updates ensure that both the device and data type are explicitly specified in a single call to .to(), improving code clarity and consistency. Additionally, the image preprocessing in kongnet.py is updated to enforce float32 precision throughout.

• Updated all instances of .to(device).type(torch.float32) to .to(device=device, dtype=torch.float32) in various infer_batch functions, ensuring that tensors are moved to the correct device and cast to float32 in a single step. This change affects the following files: efficientunet_tissue_mask_model.py [1] grandqc.py [2] hovernet.py [3] hovernetplus.py [4] kongnet.py [5] mapde.py [6] micronet.py [7] nuclick.py [8] sccnn.py [9] unet.py [10] and vanilla.py [11].

[Copilot]

Pull request overview

This PR addresses #1070 by preventing float64 tensors from being moved onto Apple Silicon’s MPS backend (which doesn’t support float64) before being cast to float32. It standardizes inference-time tensor transfers across multiple model architectures by using a single Tensor.to(device=..., dtype=torch.float32) call, and additionally ensures KongNet’s NumPy preprocessing stays in float32 to avoid producing float64 tensors upstream.

Changes:

• Update multiple infer_batch implementations to move-and-cast inputs in one call (.to(device=..., dtype=torch.float32)) to avoid MPS float64 transfer errors.
• Refactor vanilla._infer_batch input preparation to use to(..., dtype=...) instead of .to(...).type(...).
• Make KongNet.preproc explicitly float32 throughout (mean/std and scaling) to avoid returning float64 arrays.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tiatoolbox/models/architecture/vanilla.py	Uses to(device=..., dtype=torch.float32) for batch transfer/cast before permuting.
tiatoolbox/models/architecture/unet.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/sccnn.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/nuclick.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/micronet.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/mapde.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues (directly tied to #1070).
tiatoolbox/models/architecture/kongnet.py	Updates input transfer/cast in infer_batch and makes preproc float32 end-to-end.
tiatoolbox/models/architecture/hovernetplus.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/hovernet.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/grandqc.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.
tiatoolbox/models/architecture/efficientunet_tissue_mask_model.py	Updates input transfer/cast in infer_batch to avoid MPS float64 issues.

---

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.88%. Comparing base (e337bea) to head (72668a9).

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #1073   +/-   ##
========================================
  Coverage    99.88%   99.88%           
========================================
  Files           85       85           
  Lines        11625    11626    +1     
  Branches      1524     1524           
========================================
+ Hits         11612    11613    +1     
  Misses           7        7           
  Partials         6        6           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[shaneahmed]

@gozdeg Please can you review this PR?

[gozdeg]

All looks good to me