Conditional Transport Lab
Live research models

Conditional Transport Lab

New input → H200 inference → original measurement pipeline

This page sends the chosen input to the H200 server for a new model run. It is separate from the archived-result explorer. Use fictional or public research inputs; do not enter identifying patient information. Inputs are not written to disk, and returned jobs expire from memory after15 minutes.

Research demonstration. A successful run does not establish clinical reliability. The current API tunnel is session-scoped; a permanent ingress is still required for uninterrupted long-term reviewer access.

Connecting to the H200 service…

Run a fresh input

Every run samples new rows on the H200 from the preserved fit17 terminal checkpoint. The original CPU label-rank and conditional maps rank the same native continuous scores and use the benchmark TRAIN split only. A fixed seed is reproducible; changing the seed changes the draw. α blends label-rank and conditional output; intermediate blends need not lie on observed TRAIN support.

Computational path

Browser input → bounded in-memory queue → preserved model on H200 → original Python measurement or correction → new response and model identity.

Model weights may stay loaded between calls. Returned inference results are not cached. The server uses the research GPU lock and the keeper yields during model execution. Failed calls show their failure; archived outputs are not substituted.