Dev

[IamMuhammadZeeshan]

Summary

Related To

Description

Additional Notes (Optional)

[gemini-code-assist]

Code Review

This pull request transitions the storage of inference results, configurations, and metadata from plain text and JSON files to a unified HDF5 format (inference_data.hdf5), updating both FlowMC and NumPyro samplers accordingly. It also introduces an automated HTML report generation tool using a Jupyter notebook template, standardizes parameter naming across hybrid models, and adds an HDF5 repacking script. The review feedback highlights several critical issues: a potential crash in generate_marginal_probs when inspecting classes without an explicit __init__, a bug in write_to_hdf5 where JAX Arrays are not converted to NumPy arrays during dataset overwrites, an argument parsing error in h5repack.py that prevents the file argument from being matched, and a lack of robustness in the report notebook when certain datasets are missing from the HDF5 file.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

In write_to_hdf5, the elif isinstance(data, Array): block is skipped if dataset_path in f is True. This means JAX Arrays are not converted to NumPy arrays when overwriting an existing dataset, which can cause h5py to fail or behave unexpectedly. Changing elif to if ensures the conversion always runs.

Suggested change

	if data is not None:
	if dataset_path in f:
	del f[dataset_path]
	elif isinstance(data, Array):
	data = np.asarray(data)
	obj = f.create_dataset(dataset_path, data=data)
	if data is not None:
	if dataset_path in f:
	del f[dataset_path]
	if isinstance(data, Array):
	data = np.asarray(data)
	obj = f.create_dataset(dataset_path, data=data)

[gemini-code-assist]

Using inspect.signature(model_meta_cls.__init__) and then param_names.remove("self") will crash if model_meta_cls does not explicitly define __init__ (like EccentricityMattersCore), because object.__init__ has no self parameter in its signature (or raises ValueError). Using inspect.signature(model_meta_cls) is the standard, safe way to get the constructor signature without self and works correctly for all classes.

Suggested change

	param_names = list(inspect.signature(model_meta_cls.__init__).parameters.keys()) # type: ignore
	param_names.remove("self")
	param_names = list(inspect.signature(model_meta_cls).parameters.keys())

[gemini-code-assist]

Defining a positional argument "options" with nargs=argparse.REMAINDER before "file" causes "options" to consume all remaining arguments (including the file path), leaving "file" unmatched and causing the script to always fail with a required argument error. Removing the "options" argument from the parser allows parse_known_args() to correctly capture all unrecognized options in the options list while successfully parsing "file".

    parser.add_argument("file", help="Path to the HDF5 file to repack")
    args, options = parser.parse_known_args()

[gemini-code-assist]

The notebook assumes that all training/production datasets (like /acceptances/global/train, loss, /chains/train/...) are always present in the HDF5 file. If any phase was skipped or had no data, these datasets won't exist, causing read_from_hdf5 to raise a ValueError and crash the entire automated report generation. Consider checking if the datasets exist in the file before reading them.