[Bounty] Deepr PyHealth 2.0 ver.

[joshuasteier]

Contributor Information

• Name: Joshua Steier
• Contribution Type: Model

Description

Updated the Deepr model to align with PyHealth 2.0 conventions:

• Migrated from BaseEHRDataset to SampleDataset
• Replaced manual tokenization with EmbeddingModel
• Updated label handling to use get_output_size() and direct tensor access
• Added learnable gap embedding for nested sequence flattening
• Supports both 3D and 4D input tensors

Files to Review

• pyhealth/models/deepr.py
• tests/core/test_deepr.py

Testing

python -m pytest -q tests/core/test_deepr.py -s
# Result: 5 passed

[Logiquo]

Could you also provide a example notebook in the examples folder to demonstrate how to use this model? thank you.

[joshuasteier]

@jhnwu3, @LogicFan, I added an example notebook in addition to the changes above, thank you.