Skip to content

Start v0.6.3 development and record the v0.6.2 release evidence #44

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
DaoyuanLi2816 merged 1 commit into from
Aug 5, 2026
Merged

Start v0.6.3 development and record the v0.6.2 release evidence #44

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 11 additions & 3 deletions PROJECT_STATE.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ and what it printed.

Last updated: 2026-08-04.

## v0.6.2 Bridge correctness and responsive visual hardening (in progress)
## v0.6.2 Bridge correctness and responsive visual hardening

### Starting state — 2026-08-04

Expand Down Expand Up @@ -78,8 +78,16 @@ stem-specific naming, collision refusal, `--overwrite`, fault injection and conc
| docs | [`30975138352`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30975138352) green; 20 Linux screenshots uploaded as `docs-visual-screenshots` |
| pinned verl bridge | [`30975138353`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30975138353) green |
| screenshot inspection | the CI Linux screenshots were downloaded and read directly, not only trusted as green. The 390 px Alignment Lab mobile chart and metric-coverage cards render the same as the local Windows capture |
| release metadata | publication was explicitly authorized, so the version is finalized to `0.6.2`, `CHANGELOG.md` carries a dated `[0.6.2] - 2026-08-05` section with its compare link, `CITATION.cff` records `0.6.2` / `2026-08-05`, `PYPI.md` is regenerated tag-pinned to `blob/v0.6.2`, and `docs/generated/quality.json` records release `0.6.2`, status `validated`, 1652 CPU tests at 85.77% branch coverage, 5 GPU and 3 network |
| local release build | `python -m build` and `twine check` pass at `0.6.2`. These local artifacts are a sanity check only; the published distributions are built once inside the tag workflow |
| release metadata | the version was finalized to `0.6.2`, `CHANGELOG.md` carries a dated `[0.6.2] - 2026-08-05` section with its compare link, `CITATION.cff` records `0.6.2` / `2026-08-05`, `PYPI.md` is regenerated tag-pinned to `blob/v0.6.2`, and `docs/generated/quality.json` records release `0.6.2` at 1652 CPU tests, 85.77% branch coverage, 5 GPU and 3 network |
| local release build | `python -m build` and `twine check` pass at `0.6.2`. These local artifacts were a sanity check only; the published distributions are built once inside the tag workflow |
| merge | PR [#43](https://github.com/DaoyuanLi2816/mini-verl/pull/43) was squash-merged; the release commit is `bef9f0878eb3280f450aee3868b43d61f0726557` after its message was rewritten so the sole contributor is the repository owner. Synchronized-main CI, build and docs runs [`30980343307`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30980343307) / [`30980343344`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30980343344) / [`30980343293`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30980343293) are green on that exact commit |
| tag | annotated `v0.6.2` resolves to `bef9f0878eb3280f450aee3868b43d61f0726557` |
| release execution | tag run [`30980579109`](https://github.com/DaoyuanLi2816/mini-verl/actions/runs/30980579109) passed metadata validation, the full quality gate, a one-time build, OIDC Trusted Publishing with attestations, public PyPI verification, exact install and GitHub Release creation |
| published wheel | [`miniverl-0.6.2-py3-none-any.whl`](https://pypi.org/project/miniverl/0.6.2/), SHA-256 `38131c3de838b480017b2f97df3e43d53e760f272c516522a470d2812f8a3803`, 330906 bytes |
| published sdist | [`miniverl-0.6.2.tar.gz`](https://pypi.org/project/miniverl/0.6.2/), SHA-256 `bcd30863290c9c46c1e2c27c59d8ccfd3ec7e5934243f5bc86f7424d4f05333d`, 994642 bytes |
| independent public verification | the PyPI JSON API reports version `0.6.2` with a Markdown description pinned to `blob/v0.6.2`, and the integrity API exposes one Trusted Publisher attestation per distribution bound to `DaoyuanLi2816/mini-verl`, `release.yml`, environment `pypi`. The [GitHub Release](https://github.com/DaoyuanLi2816/mini-verl/releases/tag/v0.6.2) ships byte-identical distributions whose recomputed SHA-256 values match its `SHA256SUMS` and PyPI exactly. A clean Windows venv installed `miniverl==0.6.2` from `https://pypi.org/simple`, reported `miniverl 0.6.2` with torch absent, and exposes `--require-tokenizer-load`, `--scan-dataset-text` and `--sentinel` |
| version transition | `v0.6.2` is immutable and public; this state sync identifies subsequent development as `0.6.3.dev0` |
| release state | complete |

### Known limitation carried forward

Expand Down
46 changes: 23 additions & 23 deletions PYPI.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
<p align="center">
<img src="https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/v0.6.2/docs/banner.svg" alt="miniVERL — single-GPU LLM post-training" width="880">
<img src="https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/banner.svg" alt="miniVERL — single-GPU LLM post-training" width="880">
</p>

<div align="center">
Expand All @@ -8,15 +8,15 @@
[![Build](https://github.com/DaoyuanLi2816/mini-verl/actions/workflows/build.yml/badge.svg)](https://github.com/DaoyuanLi2816/mini-verl/actions/workflows/build.yml)
[![PyPI](https://img.shields.io/pypi/v/miniverl.svg)](https://pypi.org/project/miniverl/)
[![Python](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)](https://www.python.org)
[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/LICENSE)
[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/DaoyuanLi2816/mini-verl/blob/main/LICENSE)

</div>

<p align="center">
<a href="https://pypi.org/project/miniverl/"><strong>PyPI</strong></a> ·
<a href="https://daoyuanli2816.github.io/mini-verl/"><strong>Stable docs</strong></a> ·
<a href="https://daoyuanli2816.github.io/mini-verl/dev/">Development docs</a> ·
<a href="https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/README.zh-CN.md">中文</a>
<a href="https://github.com/DaoyuanLi2816/mini-verl/blob/main/README.zh-CN.md">中文</a>
</p>

**miniVERL is a local, inspectable runtime for a documented subset of
Expand Down Expand Up @@ -44,15 +44,15 @@ optimization in about 50 seconds on the measured laptop CPU. For inspection,
schemas and reports without the ML stack, use `pip install miniverl`. For CUDA
training, install the matching CUDA-enabled PyTorch wheel first, then install
`miniverl[train,cuda]`; the extra does not select a CUDA PyTorch build. See the
[single-GPU guide](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/single-gpu-guide.md).
[single-GPU guide](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/single-gpu-guide.md).

## Three paths

| Path | Start with | Concrete artifact | Next |
| --- | --- | --- | --- |
| **Align** — compare SFT, DPO, KD and OPD only when the pilot evidence supports the cost | `miniverl pilot recipes/alignment_policy_conditioned_qwen.yaml` | `alignment-card.json` | [Alignment Lab](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/alignment-lab/alignment-lab-v1.md) |
| **Distill locally** — strict OPD, shared backbones and padded trajectory updates on one CUDA GPU | `miniverl train recipes/qwen_consumer_gpu_shared.yaml --dry-run` | `config.resolved.yaml` plus a revision-pinned PEFT adapter | [Bring your own GPU](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/single-gpu-guide.md) |
| **Scale out** — import a documented profile, convert Parquet, export a bundle and run bridge checks | `miniverl bridge doctor scaleout-bundle` | `provenance/compatibility-report.json` | [Verified verl artifact bridge](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/verl-bridge.md) |
| **Align** — compare SFT, DPO, KD and OPD only when the pilot evidence supports the cost | `miniverl pilot recipes/alignment_policy_conditioned_qwen.yaml` | `alignment-card.json` | [Alignment Lab](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-lab/alignment-lab-v1.md) |
| **Distill locally** — strict OPD, shared backbones and padded trajectory updates on one CUDA GPU | `miniverl train recipes/qwen_consumer_gpu_shared.yaml --dry-run` | `config.resolved.yaml` plus a revision-pinned PEFT adapter | [Bring your own GPU](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/single-gpu-guide.md) |
| **Scale out** — import a documented profile, convert Parquet, export a bundle and run bridge checks | `miniverl bridge doctor scaleout-bundle` | `provenance/compatibility-report.json` | [Verified verl artifact bridge](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/verl-bridge.md) |

The bridge is a **verified artifact bridge**: a pinned config/data/model
parse-load smoke at miniVERL-defined compatibility Level 3. It has never run a
Expand Down Expand Up @@ -81,12 +81,12 @@ improved it; continued SFT and both OPD variants retained measured regressions.
| standard OPD | 98.6% | 97.2% | 100.0% | 76.7 s |
| verifier-gated OPD | 97.9% | 95.8% | 46.8% | 66.0 s |

![Alignment and utility deltas from the saturated SFT checkpoint; small marks are all three seeds and large marks are means](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/v0.6.2/docs/alignment-lab/delta-from-sft.svg)
![Alignment and utility deltas from the saturated SFT checkpoint; small marks are all three seeds and large marks are means](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/alignment-lab/delta-from-sft.svg)

The two sandbox safety checks tied at zero while utility still regressed.
IFEval, XSTest, HarmBench and RewardBench were **not executed**. “Preference
win rate” is a deterministic Minipolicy paired outcome, not human preference.
Read the [study, seed-level values and limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/alignment-lab/alignment-lab-v1.md).
Read the [study, seed-level values and limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-lab/alignment-lab-v1.md).

## One measured systems result

Expand All @@ -97,13 +97,13 @@ reserved memory versus 3.035 GiB for dual model, while running 10.1% slower.
All 12 preregistered equivalence comparisons passed. These are one-workload,
one-machine measurements, not promises for other GPUs.

![Measured throughput and reserved VRAM for dual-model and shared-backbone runtime cells](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/v0.6.2/docs/consumer-runtime-v1-pareto.svg)
![Measured throughput and reserved VRAM for dual-model and shared-backbone runtime cells](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/consumer-runtime-v1-pareto.svg)

[Consumer Runtime v1 methods and caveats](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/consumer-runtime-v1.md)
[Consumer Runtime v1 methods and caveats](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/consumer-runtime-v1.md)

## Compatibility boundary

![Verified local runtime, portable artifact bundle and pinned upstream smoke; distributed verl execution remains untested](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/v0.6.2/docs/verl-bridge-architecture.svg)
![Verified local runtime, portable artifact bundle and pinned upstream smoke; distributed verl execution remains untested](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/verl-bridge-architecture.svg)

The bridge targets official verl `v0.8.0` at commit `7aed6b23` and uses the
term **miniVERL-defined compatibility Level 3**. That means a checksummed
Expand All @@ -120,21 +120,21 @@ PPO/reward scaffold, not an executable continuation of miniVERL OPD semantics.

## Detailed studies and preserved negative evidence

- [RecoveryBench v1](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/recoverybench/recoverybench-v1.md): frozen-student KD
- [RecoveryBench v1](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/recoverybench/recoverybench-v1.md): frozen-student KD
outperformed much slower fresh-state OPD on the preregistered primary view;
the verifier gate remained `insufficient_evidence`.
- [Alignment Lab v1](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/alignment-lab/alignment-lab-v1.md): the starting SFT
- [Alignment Lab v1](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-lab/alignment-lab-v1.md): the starting SFT
checkpoint was at the ceiling, so no positive OPD result is claimed.
- [Calculator benchmark](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/benchmarking.md): both negative controls completed
- [Calculator benchmark](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/benchmarking.md): both negative controls completed
normally and measured 0% strict success. They were not configuration
failures. Because they used the historical ambiguous protocol-v1 prompt,
their failure cannot be attributed solely to intrinsic teacher behavior.
- [Consumer Runtime v1](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/consumer-runtime-v1.md): padded update batches and
- [Consumer Runtime v1](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/consumer-runtime-v1.md): padded update batches and
shared adapters preserve the measured one-update objective within declared
tolerances; rollout generation remains sequential.
- [Limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/limitations.md), [math](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/math.md),
[reproducibility](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/reproducibility.md) and
[compatibility policy](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/docs/compatibility.md).
- [Limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/limitations.md), [math](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/math.md),
[reproducibility](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/reproducibility.md) and
[compatibility policy](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/compatibility.md).

New runs establish tokenizer compatibility through structural identity. The
legacy behavioral fingerprint—token IDs for one fixed probe plus metadata—is
Expand All @@ -154,7 +154,7 @@ python -m pip install -e ".[dev]"
pytest -q -m "not gpu and not network"
```

Apache-2.0 licensed. See [CONTRIBUTING.md](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/CONTRIBUTING.md) and
[SECURITY.md](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/SECURITY.md). Project records: [default GPU recipe](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/recipes/qwen_consumer_gpu_calc.yaml),
[frozen calculator JSON](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/benchmarks/results/gpu-calc-hard-equal-update-v2.json),
[changelog](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/CHANGELOG.md), [citation](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/CITATION.cff) and [license](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.6.2/LICENSE).
Apache-2.0 licensed. See [CONTRIBUTING.md](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CONTRIBUTING.md) and
[SECURITY.md](https://github.com/DaoyuanLi2816/mini-verl/blob/main/SECURITY.md). Project records: [default GPU recipe](https://github.com/DaoyuanLi2816/mini-verl/blob/main/recipes/qwen_consumer_gpu_calc.yaml),
[frozen calculator JSON](https://github.com/DaoyuanLi2816/mini-verl/blob/main/benchmarks/results/gpu-calc-hard-equal-update-v2.json),
[changelog](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CHANGELOG.md), [citation](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CITATION.cff) and [license](https://github.com/DaoyuanLi2816/mini-verl/blob/main/LICENSE).
2 changes: 1 addition & 1 deletion docs/generated/quality.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"schema_version": 1,
"release": "0.6.2",
"status": "validated",
"status": "released",
"measured_commit": "58f7018c9e28bc0460856911539cfdf376354d7b",
"measured_at": "2026-08-04T22:44:33-07:00",
"cpu_non_gpu_non_network": {
Expand Down
2 changes: 1 addition & 1 deletion src/miniverl/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,6 @@

from __future__ import annotations

__version__ = "0.6.2"
__version__ = "0.6.3.dev0"

__all__ = ["__version__"]