8000 Prepare the v0.6.0 release by DaoyuanLi2816 · Pull Request #37 · DaoyuanLi2816/mini-verl · GitHub
[go: up one dir, main page]

Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,8 @@ All notable changes to miniVERL are recorded here. The format follows

## [Unreleased]

## [0.6.0] - 2026-08-03

### Added

- A compatibility Level-3 bridge for the fail-closed
Expand Down Expand Up @@ -554,7 +556,8 @@ Same-tokenizer only; one trajectory per forward pass; `swap` unavailable for
quantized models; only Qwen3 and Qwen2 architectures tested; single-seed GPU
results. The full list is in `docs/limitations.md`.

[Unreleased]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.5.0...HEAD
[Unreleased]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.6.0...HEAD
[0.6.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.5.0...v0.6.0
[0.5.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.4.0...v0.5.0
[0.4.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.3.0...v0.4.0
[0.3.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.2.6...v0.3.0
Expand Down
6 changes: 3 additions & 3 deletions CITATION.cff
Original file line number Diff line number Diff line change
Expand Up @@ -2,8 +2,8 @@ cff-version: 1.2.0
title: "miniVERL: On-policy distillation for tool-using agents on one GPU"
message: "If you use miniVERL in your work, please cite it as below."
type: software
version: 0.5.0
date-released: 2026-08-02
version: 0.6.0
date-released: 2026-08-03
license: Apache-2.0
repository-code: "https://github.com/DaoyuanLi2816/mini-verl"
url: "https://github.com/DaoyuanLi2816/mini-verl"
Expand Down Expand Up @@ -98,4 +98,4 @@ references:
url: "https://arxiv.org/abs/2603.07079"
notes: >-
Motivates recording per-token teacher entropy. Entropy-aware divergence
mixing is a roadmap item and is not implemented in miniVERL v0.5.0.
mixing is a roadmap item and is not implemented in miniVERL v0.6.0.
9 changes: 6 additions & 3 deletions PROJECT_STATE.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,9 +4,9 @@ Living build log for **miniVERL** (`mini-verl` / `miniverl` / CLI `miniverl`).
A checkbox is not evidence: every completed item names the command that was run
and what it printed.

Last updated: 2026-08-02.
Last updated: 2026-08-03.

## v0.6.0 Verified verl Bridge development state
## v0.6.0 Verified verl Bridge release candidate

| item | current state |
| --- | --- |
Expand All @@ -18,7 +18,10 @@ Last updated: 2026-08-02.
| community surface | four versioned recipe records bind existing measurements; the preference-teacher distillation entry is explicitly unmeasured and links an aligned-adapter scaffold that still requires a pinned preference-trained teacher; no external adoption is claimed |
| documentation | a static MkDocs site, bridge architecture diagram, launch article/demo/card copy/announcement/social preview and community submission guide are part of the release candidate |
| immutable evidence | the frozen calculator JSON remains required byte-identical at SHA-256 `53fc1d4d5b7adee09618d77ad62d4086ba56b78569832d6fc7c3bcd5c2695bbc`; prior tags and published adapter revisions remain immutable |
| release state | development PR in progress from `0.6.0.dev0`; publication evidence is recorded only after tag, OIDC, public install and state sync complete |
| local release gates | ruff, format, mypy, actionlint, Markdown/link and generated-artifact checks pass; the full non-GPU/non-network suite passes 1548 tests with 6 deselected at 85.53% branch coverage; the available GPU and network gates each pass 3; package build, Twine, clean Python 3.10 wheel install and exact pinned-verl smoke pass |
| integration source | Verified verl Bridge PR [#36](https://github.com/DaoyuanLi2816/mini-verl/pull/36) was squash-merged as `0d43310cca47db828b10a0e12facb33e8f0fd371`; synchronized main CI, build and docs runs `30794655713`, `30794655722` and `30794655822` are green; the latest PR bridge smoke is `30794329109` |
| public documentation | GitHub Pages deployed successfully and returned HTTP 200 at `https://daoyuanli2816.github.io/mini-verl/` with the bridge documentation visible |
| release state | development integration and post-merge validation are complete; this focused release-metadata change binds exact `0.6.0`, after which annotated-tag publication and `0.6.1.dev0` state sync remain |

## v0.5.0 One-GPU Alignment Lab release

Expand Down
92 changes: 46 additions & 46 deletions PYPI.md

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ pinned verl release for scale-out.
**Measure alignment, over-refusal, retained utility and cost before choosing
SFT, DPO or OPD.**

PyPI `v0.5.0` is the stable release; `main` is development and may be ahead.
PyPI `v0.6.0` is the stable release; `main` is development and may be ahead.

miniVERL is independent from verl and implements one verified Level-3 profile,
`single-gpu-online-distillation-v1`, against official verl `v0.8.0`. This is a
Expand Down
2 changes: 1 addition & 1 deletion README.zh-CN.md
Original file line number< F438 /th> Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@

**先同时测量对齐、过度拒绝、保留效用与成本,再选择 SFT、DPO 或 OPD。**

PyPI `v0.5.0` 是稳定发布版;`main` 是开发分支,可能领先于稳定版。
PyPI `v0.6.0` 是稳定发布版;`main` 是开发分支,可能领先于稳定版。

miniVERL 独立于 verl,并针对官方 verl `v0.8.0` 实现了一个经验证的 Level-3
配置 `single-gpu-online-distillation-v1`。它是标准产物与配置子集桥接,不是
Expand Down
20 changes: 10 additions & 10 deletions docs/generated/quality.json
Original file line number Diff line number Diff line change
@@ -1,22 +1,22 @@
{
"schema_version": 1,
"release": "0.5.0",
"status": "released",
"measured_commit": "f9dae54a203d303f6562101500343ead310d99e8",
"measured_at": "2026-08-02T22:50:12-07:00",
"release": "0.6.0",
"status": "validated",
"measured_commit": "0d43310cca47db828b10a0e12facb33e8f0fd371",
"measured_at": "2026-08-03T00:42:52-07:00",
"cpu_non_gpu_non_network": {
"passed": 1508,
"passed": 1548,
"deselected": 6,
"branch_coverage_percent": 86.13
"branch_coverage_percent": 85.53
},
"gpu": {
"passed": 5,
"deselected": 1509,
"passed": 3,
"deselected": 1551,
"hardware": "NVIDIA GeForce RTX 4080"
},
"network": {
"passed": 3,
"deselected": 1511
"deselected": 1551
},
"quality_floor": "1,500+ tests and 86%+ branch coverage at v0.5.0"
"quality_floor": "1,540+ tests and 85%+ branch coverage at v0.6.0"
}
44 changes: 44 additions & 0 deletions docs/release-checklist.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,50 @@ This is the release gate and publication record for miniVERL. A checked item
names an invariant exercised on the stated source. Publication begins only
after the exact release commit and its remote checks are green.

## v0.6.0 Verified verl Bridge release

- [x] The documented `single-gpu-online-distillation-v1` profile is pinned to
official verl `v0.8.0` commit
`7aed6b230776f963fa09509c10d9c3a767d1102c`; the installed source reports
`0.8.0.dev0` and no moving branch or mutable revision is accepted.
- [x] The fail-closed importer accepts only 14 named configuration fields; the
exported bundle supplies the six supported LoRA/reward fields and rejects
unsupported algorithm, optimizer, distributed-runtime and checkpoint
semantics instead of claiming PPO/GRPO parity.
- [x] The exact Python 3.12 smoke parses and structurally merges the official
OmegaConf, loads standard PEFT LoRA and safetensors structure, reads both
Parquet splits, imports the fail-closed reward scaffold, verifies the
tokenizer identity, privacy and 14 artifact hashes, and records
distributed execution as `not tested`.
- [x] Dataset conversion is reversible for the official prompt schema, keeps
miniVERL extensions in a checksummed sidecar and publishes Parquet through
a unique temporary file without masking the originating write error.
- [x] Five packaged community records distinguish four existing measured
artifacts from one explicit `not_measured` preference-teacher template;
no external adoption, distributed execution or new benchmark result is
claimed.
- [x] The calculator benchmark JSON remains byte-identical at SHA-256
`53fc1d4d5b7adee09618d77ad62d4086ba56b78569832d6fc7c3bcd5c2695bbc`,
and its generated SVG byte comparison passes; prior tags, negative results
and public adapter revisions remain unchanged.
- [x] Ruff check/format, mypy across 103 source files, actionlint 1.7.12,
Markdown/link checks, strict MkDocs, PyPI byte comparison and
`git diff --check` pass.
- [x] Full non-GPU/non-network suite: **1548 passed**, 6 deselected, **85.53%**
branch coverage; the available GPU/non-network and network suites each
pass **3 tests**.
- [x] The focused bridge/packaging suite passes 124 tests; wheel and sdist pass
Twine, a clean Python 3.10 wheel install passes doctor/community export,
and the wheel contains all bridge modules and five recipe records.
- [x] The architecture SVG passed native and 820-pixel README inspection; the
1280 by 640 social preview passed native inspection; GitHub Pages deploy
run `30794655822` is green and the public site returns HTTP 200.
- [x] Focused PR #36 is green and squash-merged as `0d43310`; synchronized main
CI and build runs `30794655713` and `30794655722` are green, and exact
pinned-verl PR smoke run `30794329109` is green.
- [x] Release metadata declares exact `0.6.0`; the intended annotated tag is
exactly `v0.6.0` and will use the existing OIDC-only release workflow.

## v0.5.0 One-GPU Alignment Lab release

- [x] Public preregistration revision 1.4 freezes three seeds, the common SFT
Expand Down
2 changes: 1 addition & 1 deletion src/miniverl/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,6 @@

from __future__ import annotations

__version__ = "0.6.0.dev0"
__version__ = "0.6.0"

__all__ = ["__version__"]
4 changes: 2 additions & 2 deletions tests/unit/test_packaging.py
Original file line number Diff line number Diff line change
Expand Up @@ -414,7 +414,7 @@ def test_generated_pypi_readme_is_byte_bound_and_has_only_navigable_project_link

@pytest.mark.parametrize(
("version", "ref"),
[("0.5.0", "v0.5.0"), ("0.6.0.dev0", "main")],
[("0.6.0", "v0.6.0"), ("0.6.1.dev0", "main")],
)
def test_pypi_generator_rewrites_nested_images_and_every_project_target(
tmp_path: Path,
Expand Down Expand Up @@ -474,7 +474,7 @@ def test_release_quality_has_one_version_bound_machine_readable_record() -> None
(REPO_ROOT / "docs" / "generated" / "quality.json").read_text(encoding="utf-8")
)
assert record["schema_version"] == 1
assert record["quality_floor"] == "1,500+ tests and 86%+ branch coverage at v0.5.0"
assert record["quality_floor"] == "1,540+ tests and 85%+ branch coverage at v0.6.0"
if ".dev" in miniverl.__version__:
assert record["status"] in {"candidate", "released"}
if record["status"] == "candidate":
Expand Down
0