Add NestedMultiHotProcessor and vectorise HALO's visit encoding - #1233

Merged

jhnwu3 merged 9 commits into

sunlabuiuc:masterfrom

janezdu:multihot-ehrgen

Sep 12, 2026

Commits on Sep 9, 2026

Add NestedMultiHotProcessor and vectorise HALO's visit encoding

janezdu
and
claude
committed
d047035
Satisfy PR contribution rules: lint, docstring examples, update HALO example

janezdu
and
claude
committed
db31122

Commits on Sep 12, 2026

Drop the encode-equivalence test, keep the truncation coverage

janezdu
and
claude
committed
b7e563e
Share the code vocabulary, and fix GPT2/PromptEHR under the new encoding

janezdu
and
claude
committed
57aaa2b
Give each generator family its own EHR-generation task

janezdu
and
claude
committed
83ac1f6
Revert the shared code-vocabulary mixin

janezdu
and
claude
committed
79c4c3f
Drop the unused multi-hot inverse; use the task helpers in the example

janezdu
and
claude
committed
d947b63
Cover the task extraction and the code-set pooling

janezdu
and
claude
committed
bd683b2
Review: flatten the generation tasks, name the dataset, add examples

janezdu
and
claude
committed
db9e4c4