bax_algorithms

Algorithms for Bayesian Algorithm Execution (BAX) with Xopt.

Remote BAX emittance examples

These examples separate experiment evaluation from generator computation:

• Client Host: runs Jupyter, the Xopt application, and the evaluator.
• Remote Compute Backend: runs the Xopt HTTP server and performs BAX model fitting and candidate generation on CPU or GPU.

The client connects to the backend through an authenticated HTTP generator session. For the development testbed, HTTP traffic is carried through an SSH tunnel.

Notebook	Remote generator compute	Authentication
examples/http_generator_emittance.ipynb	CPU	Bearer token
examples/http_generator_emittance_gpu.ipynb	GPU	Bearer token
examples/http_generator_emittance_gpu_authenticated.ipynb	GPU	Bearer token (explicit example)

Quick start: configure access to a running Remote Compute Backend, establish an SSH tunnel, set XOPT_REMOTE_URL, and run one of the notebooks from a Python environment with Xopt and the example dependencies installed. Credentials are loaded from a local file, never from notebook source.

See Remote BAX Emittance Guide for setup, commands, regression testing, and troubleshooting.

Regression testing

Run notebook execution tests from the repository root:

python scripts/test_http_notebooks.py \
  examples/http_generator_emittance.ipynb \
  examples/http_generator_emittance_gpu.ipynb \
  examples/http_generator_emittance_gpu_authenticated.ipynb

The runner reports per-cell progress and timings and does not overwrite the notebook files. It requires a working Remote Compute Backend, SSH tunnel, and authentication token; these are integration tests, not offline unit tests.