v0.2.1

What's Changed

• fix: Aggregate mode 1 neural statistics per core by @Rob-Hughes-Arm
• feat: Use graph compiler directly on TOSA models by @integratedintegral21
• Check tosa version for compatibility checks by @integratedintegral21
• fix: Share API label parsing across estimator table shapes by @Rob-Hughes-Arm
• Skip trivial ops compat by @integratedintegral21
• chore: Remove pytorch-e2e dependency group by @integratedintegral21

Full Changelog: v0.2.0...v0.2.1

v0.2.1rc1

What's Changed

• fix: Aggregate mode 1 neural statistics per core by @Rob-Hughes-Arm
• feat: Use graph compiler directly on TOSA models by @integratedintegral21
• Check tosa version for compatibility checks by @integratedintegral21
• fix: Share API label parsing across estimator table shapes by @Rob-Hughes-Arm
• Skip trivial ops compat by @integratedintegral21
• chore: Remove pytorch-e2e dependency group by @integratedintegral21

Full Changelog: v0.2.0...v0.2.1rc1

v0.2.0

What's Changed

• feat: Add NX Performance estimator EULA and artifact by @IsabellaARM
• feat: Add simple EULA display and confirmation flow by @IsabellaARM
• feat: Log crossing of internal boundaries by @oluwatamilore-arm
• fix: Share API label recovery across estimation and profiling by @Rob-Hughes-Arm
• Preserve shared source provenance in NX performance by @Rob-Hughes-Arm
• fix: Allow compatible MLIA Core 0.12 updates by @max-shvetsov

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

• feat: Add contribution guidelines for public release by @max-shvetsov
• fix: Allow vault secrets fetch outside of fork PRs. by @max-shvetsov
• ci: Read uv default index from Vault by @max-shvetsov
• feat: Rename MLIA Tensorflow Lite converter to LiteRT by @IsabellaARM
• feat: Introduce vault secrets by @oluwatamilore-arm
• chore: Fix mlia core version to 0.12.1 by @integratedintegral21
• feat: Publish versioned release documentation by @max-shvetsov
• ci: Codex PR review by @jslater8
• chore: Remove events and handlers usage by @integratedintegral21
• refactor: Use new Entity definition by @integratedintegral21
• fix(nx): Support NX estimator across Linux and Windows by @Rob-Hughes-Arm
• fix(compatibility): Show operator types in entity names by @Rob-Hughes-Arm
• ci: Test Linux workflows with Python 3.12 by @Rob-Hughes-Arm
• build: Bundle VGF Python bindings from source by @Rob-Hughes-Arm
• fix(nx): Align default system cache configuration by @Rob-Hughes-Arm
• feat(output): Add canonical NX entity provenance by @Rob-Hughes-Arm
• feat(profiling): Add structured capture ingestion and measured analysis by @Rob-Hughes-Arm
• fix(profiling): Handle capture metadata and zero counters by @jslater8
• fix: Show NX placement for chain entities in performance runs by @integratedintegral21
• fix: Use uppercase NX and EE in compatibility runs by @integratedintegral21
• fix: PyTorch e2e release tests by @integratedintegral21
• ci: Add syncing with the public repo by @integratedintegral21
• fix(compatibility): Preserve truthful operator context by @Rob-Hughes-Arm
• docs: Align Neural Technology guidance with runtime by @Rob-Hughes-Arm
• feat: Move to Open Source License except for Performance Estimator by @IsabellaARM
• fix: resolve PyTorch from CPU index by @max-shvetsov
• ci: Publish documentation to GitHub Pages by @IsabellaARM
• fix: Replace relative metadata links with valid web urls by @integratedintegral21
• ci: Sync release notes to public repositories by @IsabellaARM
• docs: Update documentation and notebook wording by @IsabellaARM
• ci: Add TOSA models for end-to-end tests by @integratedintegral21
• docs: Add more info about Performance Estimator output metrics by @IsabellaARM
• ci: Refactor e2e by @integratedintegral21
• ci: align release checks and public synchronization by @max-shvetsov
• fix: configure Black Duck token through Vault mappings by @max-shvetsov
• fix: read release wheel metadata without shell pipes by @max-shvetsov
• fix: reject unshaped VGF before running GCPE by @hamars01
• feat: Support executorch 1.4 by @integratedintegral21
• fix: pin dependencies compatible with MLIA 0.12.2 by @max-shvetsov
• chore: bump mlia-test by @integratedintegral21
• ci: Read test cases from mlia-test-config by @integratedintegral21
• fix: Set mlia-test-config ref to main in release.yml by @integratedintegral21
• fix: Do not run e2e on PRs by @integratedintegral21
• fix: enforce Ubuntu 24.04 baseline for Linux wheels by @max-shvetsov

New Contributors

• @oluwatamilore-arm made their first contribution
• @jslater8 made their first contribution
• @Rob-Hughes-Arm made their first contribution
• @hamars01 made their first contribution

Full Changelog: v0.1.0...v0.1.1

v0.1.1rc5

What's Changed

• chore: bump mlia-test by @integratedintegral21
• ci: Read test cases from mlia-test-config by @integratedintegral21
• fix: Set mlia-test-config ref to main in release.yml by @integratedintegral21
• fix: Do not run e2e on PRs by @integratedintegral21
• fix: enforce Ubuntu 24.04 baseline for Linux wheels by @max-shvetsov

Full Changelog: v0.1.1rc4...v0.1.1rc5

v0.1.1rc4

What's Changed

• feat: Support executorch 1.4 by @integratedintegral21
• fix: pin dependencies compatible with MLIA 0.12.2 by @max-shvetsov

Full Changelog: v0.1.1rc3...v0.1.1rc4