v1.0.x series (current stable)¶

v1.0.0 (29th September 2025)¶

This major release is a milestone with minor changes compared to v0.31.0.

Warning

The eradiate.run() now returns values more consistently according to the processed measures. In rare cases, this can be a breaking change. See below for details.

Changed¶

⚠️ 🖥️ The Experiment.contexts(), Experiment.context_init() and Experiment._context_kwargs() properties have been turned into regular methods, effectively allowing to pass additional contextual information when calling them (PR515).

Fixed¶

Improved kernel context generation to avoid unnecessary processing in some multi-measure setups (PR515).
⚠️ Improved eradiate.run() return value consistency (PR516).

v1.0.1 (14th October 2025)¶

This minor release contains 2 critical fixes and 2 minor improvements.

Added¶

Added a Matplotlib style sheet (PR523).
Enabled all of Mitsuba’s reconstruction filters (PR524).

Fixed¶

Fixed post-processing pipeline raising with Hamilton v1.89 (PR522).
Fixed broken asset manager on Windows (025d75).

v1.0.2 (9th March 2026)¶

This minor release fixes issues with critical dependencies, hopefully allowing users to deploy it without issues in a normal setup.

Fixed¶

Ensured compatibility with recent versions of Pinttrs and cachetools (PR556).