miniVERL v0.16.0

What's Changed

• chore: synchronize v0.15.0 publication and 0.15.1.dev0 by @DaoyuanLi2816 in #122
• v0.16.0: native verl Hydra input and deterministic sampling by @DaoyuanLi2816 in #123

Full Changelog: v0.15.0...v0.16.0

miniVERL v0.15.0

What's Changed

• Synchronize v0.14.0 publication and begin 0.14.1 development by @DaoyuanLi2816 in #119
• v0.15.0: run resolved verl configs directly on one GPU by @DaoyuanLi2816 in #120
• Include v0.15 direct-workflow evidence in release assets by @DaoyuanLi2816 in #121

Full Changelog: v0.14.0...v0.15.0

miniVERL v0.14.0

What's Changed

• Advance development after v0.13.0 release by @DaoyuanLi2816 in #117
• Productize local PPO and GRPO workflows for v0.14.0 by @DaoyuanLi2816 in #118

Full Changelog: v0.13.0...v0.14.0

miniVERL v0.13.0

What's Changed

• Advance development after v0.12.0 release by @DaoyuanLi2816 in #114
• Release v0.13.0: executable single-GPU PPO by @DaoyuanLi2816 in #115
• Fix v0.13 qualification release assets by @DaoyuanLi2816 in #116

Full Changelog: v0.12.0...v0.13.0

miniVERL v0.12.0

What's Changed

• Advance development to 0.11.1.dev0 by @DaoyuanLi2816 in #104
• Release v0.12.0: single-GPU verl v0.9 RL by @DaoyuanLi2816 in #111
• Fix target-length qualification metadata by @DaoyuanLi2816 in #112
• Finalize v0.12 release gate records by @DaoyuanLi2816 in #113

Full Changelog: v0.11.0...v0.12.0

miniVERL v0.11.0

What's Changed

• chore: advance development after v0.10.1 by @DaoyuanLi2816 in #87
• Polish v1 readiness and release evidence layout by @DaoyuanLi2816 in #88
• Reframe documentation around product workflows by @DaoyuanLi2816 in #90
• Advance development to 0.11.0.dev0 and freeze rollout baseline by @DaoyuanLi2816 in #91
• Add typed Rollout Runtime v2 and batched KV-cache HF backend by @DaoyuanLi2816 in #92
• Add transactional grouped rollout semantics by @DaoyuanLi2816 in #93
• Add explicit rewarded OPD profile by @DaoyuanLi2816 in #94
• Add the measured single-GPU vLLM rollout backend by @DaoyuanLi2816 in #95
• Publish RTX 4080 rollout and competitor evidence by @DaoyuanLi2816 in #96
• Qualify Rollout Runtime v2 for v0.11.0 by @DaoyuanLi2816 in #98
• Require full WSL2 qualification for v0.11 by @DaoyuanLi2816 in #99
• Prepare v0.11.0 release metadata by @DaoyuanLi2816 in #100
• Keep GPU qualification inside its virtualenv by @DaoyuanLi2816 in #101
• Isolate GPU qualification work from the checkout by @DaoyuanLi2816 in #102
• Fix v0.11 release evidence environment binding by @DaoyuanLi2816 in #103

Full Changelog: v0.10.1...v0.11.0

miniVERL v0.10.1

What's Changed

• Advance development to 0.10.1.dev0 by @DaoyuanLi2816 in #84
• Harden exact-SHA GPU qualification and maintainer release gates by @DaoyuanLi2816 in #85
• Prepare v0.10.1 release metadata by @DaoyuanLi2816 in #86

Full Changelog: v0.10.0...v0.10.1

Artifact guide

Canonical distributions:

• miniverl-0.10.1-py3-none-any.whl
• miniverl-0.10.1.tar.gz
• SHA256SUMS

Canonical release provenance:

• candidate-manifest.json
• qualification.json
• qualification-full-SHA256SUMS

Primary full-qualification results:

• miniverl-0.10.1-release-smoke-record.json
• miniverl-0.10.1-full-direct-result.json
• miniverl-0.10.1-full-pg-k1-result.json
• miniverl-0.10.1-full-smollm2-result.json

Some additional v0.10.1 evidence assets retain duplicated filename suffixes from the original flattening step. They are preserved byte-for-byte as historical release evidence; future releases use the canonical asset layout documented in the repository.

miniVERL v0.10.0

What's Changed

• Advance development to 0.10.0.dev0 by @DaoyuanLi2816 in #78
• Add versioned compatibility profile registry by @DaoyuanLi2816 in #79
• Add measured sampled-k1 OPD profile by @DaoyuanLi2816 in #80
• Measure the full SmolLM2 OPD recipe by @DaoyuanLi2816 in #81
• Add portable hardware evidence records by @DaoyuanLi2816 in #82
• Release miniVERL v0.10.0 by @DaoyuanLi2816 in #83

Full Changelog: v0.9.1...v0.10.0

miniVERL v0.9.1

What's Changed

• Advance development to 0.9.1.dev0 by @DaoyuanLi2816 in #74
• Repair verl OPD semantic contracts by @DaoyuanLi2816 in #75
• Align verl OPD documentation with runtime truth by @DaoyuanLi2816 in #76
• Release v0.9.1 by @DaoyuanLi2816 in #77

Full Changelog: v0.9.0...v0.9.1

miniVERL v0.9.0

What's Changed

• Advance development to 0.9.0.dev0 by @DaoyuanLi2816 in #67
• Add verl-style override workflow by @DaoyuanLi2816 in #68
• Bind runs to immutable execution plans by @DaoyuanLi2816 in #69
• Add bounded CUDA hardware probe by @DaoyuanLi2816 in #70
• Materialize launchable verl OPD bundles by @DaoyuanLi2816 in #71
• Publish the realistic RTX 4080 OPD workload by @DaoyuanLi2816 in #72
• Prepare the v0.9.0 release by @DaoyuanLi2816 in #73

Full Changelog: v0.8.1...v0.9.0