chore(meta): post-cutover URL sweep — lusoris/vmaf → VMAFx/vmafx

.claude/agents/pr-body-checker.md

@@ -41,8 +41,8 @@ the legacy single `CHANGELOG.md`); same for "Rebase note" →
 ## How to invoke
 
 ```bash
-gh pr view <num> --repo lusoris/vmaf --json body -q .body \
-  | PR_BODY="$(gh pr view <num> --repo lusoris/vmaf --json body -q .body)" \
+gh pr view <num> --repo VMAFx/vmafx --json body -q .body \
+  | PR_BODY="$(gh pr view <num> --repo VMAFx/vmafx --json body -q .body)" \
     bash scripts/ci/validate-pr-body.sh
 ```
 

.github/ISSUE_TEMPLATE/config.yml

@@ -1,10 +1,10 @@
 blank_issues_enabled: false
 contact_links:
   - name: Security vulnerability (private)
-    url: https://github.com/lusoris/vmaf/security/advisories/new
+    url: https://github.com/VMAFx/vmafx/security/advisories/new
     about: Report a security issue privately. Do NOT file a public issue.
   - name: Question / discussion
-    url: https://github.com/lusoris/vmaf/discussions
+    url: https://github.com/VMAFx/vmafx/discussions
     about: Open-ended questions and design discussions.
   - name: Upstream Netflix/vmaf
     url: https://github.com/Netflix/vmaf/issues

.github/workflows/docker-publish-production.yml

@@ -5,12 +5,12 @@
 #   - workflow_dispatch for manual runs
 #
 # Produces:
-#   ghcr.io/lusoris/vmafx:<tag>           (cpu, amd64+arm64)
-#   ghcr.io/lusoris/vmafx:latest          (same)
-#   ghcr.io/lusoris/vmafx:<tag>-cuda12    (amd64 only — CUDA not portable to arm64)
-#   ghcr.io/lusoris/vmafx:<tag>-rocm6     (amd64 only)
-#   ghcr.io/lusoris/vmafx:<tag>-oneapi2026 (amd64 only)
-#   ghcr.io/lusoris/vmafx:<tag>-vulkan    (amd64+arm64)
+#   ghcr.io/vmafx/vmafx:<tag>           (cpu, amd64+arm64)
+#   ghcr.io/vmafx/vmafx:latest          (same)
+#   ghcr.io/vmafx/vmafx:<tag>-cuda12    (amd64 only — CUDA not portable to arm64)
+#   ghcr.io/vmafx/vmafx:<tag>-rocm6     (amd64 only)
+#   ghcr.io/vmafx/vmafx:<tag>-oneapi2026 (amd64 only)
+#   ghcr.io/vmafx/vmafx:<tag>-vulkan    (amd64+arm64)
 #
 # Post-push steps: cosign keyless sign + syft SBOM (CycloneDX JSON) via cosign attest.
 # ADR-0698: vmafx-production-dockerfile
@@ -34,7 +34,7 @@ permissions:
 
 env:
   REGISTRY: ghcr.io
-  IMAGE_NAME: lusoris/vmafx
+  IMAGE_NAME: vmafx/vmafx
 
 jobs:
   # -----------------------------------------------------------------------

CHANGELOG.md

@@ -304,7 +304,7 @@
   [`docs/research/0006-tinyai-ptq-accuracy-targets.md`](docs/research/0006-tinyai-ptq-accuracy-targets.md)
   §4 now references the actual shipped `vmaf_tiny_v1_medium.onnx`
   checkpoint (landed by [PR
-  #158](https://github.com/lusoris/vmaf/pull/158)) instead of the
+  #158](https://github.com/VMAFx/vmafx/pull/158)) instead of the
   fictional `tiny-vmaf-v2` prototype name; the digest's QAT
   cost/budget framing is unchanged.  (c)
   [`python/vmaf/routine.py`](python/vmaf/routine.py) — both
@@ -2055,7 +2055,7 @@
   SIMD coverage matrix and `float_ms_ssim` <176×176 minimum
   documented in [`docs/metrics/features.md`](docs/metrics/features.md).
 - **Tracked `docs/state.md` + bug-status hygiene rule** (fork-local):
-  closes [Issue #20](https://github.com/lusoris/vmaf/issues/20) and
+  closes [Issue #20](https://github.com/VMAFx/vmafx/issues/20) and
   backlog item T7-1. New tracked file [`docs/state.md`](docs/state.md)
   is the canonical in-tree register of bug status (Open / Recently
   closed / Confirmed not-affected / Deferred). New CLAUDE.md §12

LICENSES/Apache-2.0-u2netp.txt

@@ -4,7 +4,7 @@ Third-party license bundle for the fork-local mirror of U-2-Net's
 Per ADR-0325 (`docs/adr/0325-u2netp-fork-mirror-scaffold.md`),
 the fork redistributes the upstream `xuebinqin/U-2-Net` `u2netp`
 checkpoint (and/or its ONNX rewrap) as a GitHub Release asset
-attached to the lusoris/vmaf repository. This file is the
+attached to the VMAFx/vmafx repository. This file is the
 Apache-2.0 §4 (a) + (c) compliance receipt — it travels with the
 binary in the same release.
 
@@ -261,8 +261,8 @@ END OF APACHE-2.0 LICENSE TEXT
 The above licence text is the canonical Apache-2.0 from
 <https://www.apache.org/licenses/LICENSE-2.0.txt> and applies to
 the redistributed `u2netp` checkpoint and any ONNX rewrap thereof
-attached to lusoris/vmaf GitHub Releases. The fork's own code
-under <https://github.com/lusoris/vmaf> is BSD-3-Clause-Plus-Patent
+attached to VMAFx/vmafx GitHub Releases. The fork's own code
+under <https://github.com/VMAFx/vmafx> is BSD-3-Clause-Plus-Patent
 (see the top-level `LICENSE` file); the two licences co-exist
 because the fork redistributes the upstream Apache-2.0 weights as
 a separately-attached release artefact, not as bundled fork

README.md

@@ -1,13 +1,13 @@
 # VMAF — Lusoris Fork
 
-[![Tests & Quality Gates](https://github.com/lusoris/vmaf/actions/workflows/tests-and-quality-gates.yml/badge.svg)](https://github.com/lusoris/vmaf/actions/workflows/tests-and-quality-gates.yml)
-[![Lint & Format](https://github.com/lusoris/vmaf/actions/workflows/lint-and-format.yml/badge.svg)](https://github.com/lusoris/vmaf/actions/workflows/lint-and-format.yml)
-[![Security Scans](https://github.com/lusoris/vmaf/actions/workflows/security-scans.yml/badge.svg)](https://github.com/lusoris/vmaf/actions/workflows/security-scans.yml)
-[![libvmaf Build Matrix](https://github.com/lusoris/vmaf/actions/workflows/libvmaf-build-matrix.yml/badge.svg)](https://github.com/lusoris/vmaf/actions/workflows/libvmaf-build-matrix.yml)
-[![FFmpeg Integration](https://github.com/lusoris/vmaf/actions/workflows/ffmpeg-integration.yml/badge.svg)](https://github.com/lusoris/vmaf/actions/workflows/ffmpeg-integration.yml)
+[![Tests & Quality Gates](https://github.com/VMAFx/vmafx/actions/workflows/tests-and-quality-gates.yml/badge.svg)](https://github.com/VMAFx/vmafx/actions/workflows/tests-and-quality-gates.yml)
+[![Lint & Format](https://github.com/VMAFx/vmafx/actions/workflows/lint-and-format.yml/badge.svg)](https://github.com/VMAFx/vmafx/actions/workflows/lint-and-format.yml)
+[![Security Scans](https://github.com/VMAFx/vmafx/actions/workflows/security-scans.yml/badge.svg)](https://github.com/VMAFx/vmafx/actions/workflows/security-scans.yml)
+[![libvmaf Build Matrix](https://github.com/VMAFx/vmafx/actions/workflows/libvmaf-build-matrix.yml/badge.svg)](https://github.com/VMAFx/vmafx/actions/workflows/libvmaf-build-matrix.yml)
+[![FFmpeg Integration](https://github.com/VMAFx/vmafx/actions/workflows/ffmpeg-integration.yml/badge.svg)](https://github.com/VMAFx/vmafx/actions/workflows/ffmpeg-integration.yml)
 [![License: BSD-3-Clause+Patent](https://img.shields.io/badge/License-BSD--3--Clause--Plus--Patent-blue.svg)](LICENSE)
 [![Conventional Commits](https://img.shields.io/badge/Conventional%20Commits-1.0.0-%23FE5196?logo=conventionalcommits)](https://www.conventionalcommits.org)
-[![OpenSSF Scorecard](https://api.securityscorecards.dev/projects/github.com/lusoris/vmaf/badge)](https://securityscorecards.dev/viewer/?uri=github.com/lusoris/vmaf)
+[![OpenSSF Scorecard](https://api.securityscorecards.dev/projects/github.com/VMAFx/vmafx/badge)](https://securityscorecards.dev/viewer/?uri=github.com/VMAFx/vmafx)
 [![ko-fi](https://img.shields.io/badge/ko--fi-support%20lusoris-ff5e5b?logo=kofi&logoColor=white)](https://ko-fi.com/lusoris)
 
 **A GPU-accelerated, full-precision, signed-release fork of

SECURITY.md

@@ -20,7 +20,7 @@ the current `master` and the latest two tagged releases.
 **Please do not open a public issue for security problems.**
 
 Use GitHub's private vulnerability-reporting flow:
-`https://github.com/lusoris/vmaf/security/advisories/new`
+`https://github.com/VMAFx/vmafx/security/advisories/new`
 
 If the issue is in code inherited from upstream, we will coordinate
 disclosure with Netflix/vmaf maintainers.

SUPPORT.md

@@ -13,7 +13,7 @@ We track upstream but don't duplicate its discussion threads.
 ## Is it about this fork specifically?
 
 - **GPU backends (CUDA / SYCL / Vulkan; HIP planned)** —
-  [open a bug report](https://github.com/lusoris/vmaf/issues/new/choose)
+  [open a bug report](https://github.com/VMAFx/vmafx/issues/new/choose)
 - **SIMD paths (AVX2 / AVX-512 / NEON)** — same
 - **`--precision` flag / IEEE-754 output** — same
 - **Tiny-AI surface / ONNX Runtime integration** — same
@@ -42,7 +42,7 @@ Questions that include all five usually get triaged within a few days.
 ## Chat / real-time
 
 None currently. If there's demand we'll open a Discord. Watch
-[the repo](https://github.com/lusoris/vmaf) for announcements.
+[the repo](https://github.com/VMAFx/vmafx) for announcements.
 
 ## Sponsoring
 

changelog.d/_pre_fragment_legacy.md

@@ -194,7 +194,7 @@
   [`docs/research/0006-tinyai-ptq-accuracy-targets.md`](docs/research/0006-tinyai-ptq-accuracy-targets.md)
   §4 now references the actual shipped `vmaf_tiny_v1_medium.onnx`
   checkpoint (landed by [PR
-  #158](https://github.com/lusoris/vmaf/pull/158)) instead of the
+  #158](https://github.com/VMAFx/vmafx/pull/158)) instead of the
   fictional `tiny-vmaf-v2` prototype name; the digest's QAT
   cost/budget framing is unchanged.  (c)
   [`python/vmaf/routine.py`](python/vmaf/routine.py) — both
@@ -1945,7 +1945,7 @@
   SIMD coverage matrix and `float_ms_ssim` <176×176 minimum
   documented in [`docs/metrics/features.md`](docs/metrics/features.md).
 - **Tracked `docs/state.md` + bug-status hygiene rule** (fork-local):
-  closes [Issue #20](https://github.com/lusoris/vmaf/issues/20) and
+  closes [Issue #20](https://github.com/VMAFx/vmafx/issues/20) and
   backlog item T7-1. New tracked file [`docs/state.md`](docs/state.md)
   is the canonical in-tree register of bug status (Open / Recently
   closed / Confirmed not-affected / Deferred). New CLAUDE.md §12

changelog.d/changed/post-cutover-url-sweep.md

@@ -0,0 +1,5 @@
+- All in-tree references to `lusoris/vmaf` (GitHub URLs, repository slugs, GHCR
+  registry paths, OCI labels) updated to `VMAFx/vmafx` following the GitHub
+  organization cutover. GHCR image paths use lowercase `vmafx/vmafx` per OCI
+  registry convention. 113 files updated, 279 occurrences replaced. PR:
+  chore/post-cutover-url-sweep.

changelog.d/fixed/avx512-float-conv-width-alignment-emergency.md

@@ -1 +1 @@
-- **CRITICAL**: AVX-512 float convolution dispatch from [#1261](https://github.com/lusoris/vmaf/pull/1261) (ADR-0504) reads past the row buffer when row width isn't a multiple of 16 floats (64 bytes). Surfaces as `munmap_chunk(): invalid pointer` / SIGABRT in `speed_temporal` which calls `vif_filter1d_s` at downscaled widths like 45 (= 360 >> 3) on portrait sources. Guard the three AVX-512 dispatch sites in `core/src/feature/vif_tools.c` with `(w % 16) == 0` — narrower rows fall through to the AVX2 8-wide path. Caught by the CHUG re-extract failing with SIGABRT on every portrait clip after the dispatch landed in master.
+- **CRITICAL**: AVX-512 float convolution dispatch from [#1261](https://github.com/VMAFx/vmafx/pull/1261) (ADR-0504) reads past the row buffer when row width isn't a multiple of 16 floats (64 bytes). Surfaces as `munmap_chunk(): invalid pointer` / SIGABRT in `speed_temporal` which calls `vif_filter1d_s` at downscaled widths like 45 (= 360 >> 3) on portrait sources. Guard the three AVX-512 dispatch sites in `core/src/feature/vif_tools.c` with `(w % 16) == 0` — narrower rows fall through to the AVX2 8-wide path. Caught by the CHUG re-extract failing with SIGABRT on every portrait clip after the dispatch landed in master.

changelog.d/fixed/cambi-cuda-host-preprocessing.md

@@ -3,7 +3,7 @@
 `integer_cambi_cuda.c::submit_fex_cuda` called `vmaf_cambi_preprocessing(dist_pic, ...)`
 directly on the CUDA picture. The CUDA picture's `data[0]` is a device pointer; reading
 it on the host caused a segfault (SIGSEGV, exit code 139) on every `--feature cambi_cuda`
-invocation. Fixes [Issue #857](https://github.com/lusoris/vmaf/issues/857).
+invocation. Fixes [Issue #857](https://github.com/VMAFx/vmafx/issues/857).
 
 **Root cause:** `vmaf_cambi_preprocessing` internally calls
 `decimate_generic_uint8_and_convert_to_10b` which dereferences `pic->data[0]` row-by-row

changelog.d/fixed/extractor-hdr-and-hfr-feature-options.md

@@ -1,7 +1,7 @@
 - The K150K/CHUG feature extractor (`ai/scripts/extract_k150k_features.py`)
   now passes HDR-aware and HFR-aware per-feature options to libvmaf
   ([ADR-0446](../docs/adr/0446-extractor-hdr-and-hfr-feature-options.md)),
-  closing [Issue #837](https://github.com/lusoris/vmaf/issues/837) +
+  closing [Issue #837](https://github.com/VMAFx/vmafx/issues/837) +
   the parallel CAMBI HDR-EOTF gap surfaced by lawrence's review of the
   in-flight CHUG extraction. Per-clip ffprobe now surfaces
   `color_primaries`/`color_transfer`/`color_space`; HDR sources

changelog.d/fixed/test-yuv-fixtures-md5.md

@@ -1 +1 @@
-- Test YUV fixture provisioner: `scripts/test/fetch-test-yuvs.sh` downloads `src01_hrc0[0-1]_576x324.yuv` from `Netflix/vmaf_resource` and md5-verifies them. Reverts the [#1237](https://github.com/lusoris/vmaf/pull/1237) ADM2 golden override, which was based on output from stale local fixture content. See [ADR-0493](docs/adr/0493-test-yuv-fixture-md5-verification.md) and [docs/development/test-fixtures.md](docs/development/test-fixtures.md).
+- Test YUV fixture provisioner: `scripts/test/fetch-test-yuvs.sh` downloads `src01_hrc0[0-1]_576x324.yuv` from `Netflix/vmaf_resource` and md5-verifies them. Reverts the [#1237](https://github.com/VMAFx/vmafx/pull/1237) ADM2 golden override, which was based on output from stale local fixture content. See [ADR-0493](docs/adr/0493-test-yuv-fixture-md5-verification.md) and [docs/development/test-fixtures.md](docs/development/test-fixtures.md).

changelog.d/fixed/tiny-registry-sha256-after-1226.md

@@ -1 +1 @@
-- `model/tiny/registry.json`: refresh `vmaf_tiny_v1_medium` sha256 to match the inlined-external-data file shipped by [#1226](https://github.com/lusoris/vmaf/pull/1226). The repack changed the file's bytes (and therefore its sha) but the registry entry was not updated in the same PR, so `meson test --suite=dnn` failed locally and would fail in CI on `test_registry`.
+- `model/tiny/registry.json`: refresh `vmaf_tiny_v1_medium` sha256 to match the inlined-external-data file shipped by [#1226](https://github.com/VMAFx/vmafx/pull/1226). The repack changed the file's bytes (and therefore its sha) but the registry entry was not updated in the same PR, so `meson test --suite=dnn` failed locally and would fail in CI on `test_registry`.

core/src/dnn/model_loader.c

@@ -1094,13 +1094,13 @@ int vmaf_dnn_verify_signature(const char *onnx_path, const char *registry_path)
 
     /* posix_spawnp argv. The certificate-identity-regexp + oidc-issuer
      * mirror docs/ai/security.md; they pin verification to
-     * lusoris/vmaf's supply-chain workflow identity. */
+     * VMAFx/vmafx's supply-chain workflow identity. */
     char *argv[] = {
         (char *)"cosign",
         (char *)"verify-blob",
         bundle_arg,
         (char *)"--certificate-identity-regexp",
-        (char *)"https://github.com/lusoris/vmaf/.github/workflows/.+",
+        (char *)"https://github.com/VMAFx/vmafx/.github/workflows/.+",
         (char *)"--certificate-oidc-issuer",
         (char *)"https://token.actions.githubusercontent.com",
         (char *)onnx_path,

deploy/helm/vmafx/Chart.yaml

@@ -8,9 +8,9 @@ description: >
 type: application
 version: 0.1.0
 appVersion: "3.0.0"
-home: https://github.com/lusoris/vmaf
+home: https://github.com/VMAFx/vmafx
 sources:
-  - https://github.com/lusoris/vmaf
+  - https://github.com/VMAFx/vmafx
 keywords:
   - vmaf
   - video-quality

deploy/helm/vmafx/templates/NOTES.txt

@@ -25,7 +25,7 @@ To scale:
 
 To roll out a new image:
   kubectl set image -n {{ .Release.Namespace }} deployment/{{ include "vmafx.fullname" . }} \
-    {{ include "vmafx.name" . }}=ghcr.io/lusoris/vmafx:<new-tag>
+    {{ include "vmafx.name" . }}=ghcr.io/vmafx/vmafx:<new-tag>
 
 {{- else if eq .Values.workload "Job" }}
 ----------------------------------------------------------------------

deploy/helm/vmafx/values.yaml

@@ -11,7 +11,7 @@
 # Image
 # ---------------------------------------------------------------------------
 image:
-  repository: ghcr.io/lusoris/vmafx
+  repository: ghcr.io/vmafx/vmafx
   tag: ""           # Defaults to .Chart.AppVersion when empty.
   pullPolicy: IfNotPresent
 

dev/Containerfile

@@ -41,7 +41,7 @@ FROM ubuntu:26.04@sha256:f3d28607ddd78734bb7f71f117f3c6706c666b8b76cbff7c9ff6e57
 
 LABEL org.opencontainers.image.title="vmaf-dev-mcp-build-deps"
 LABEL org.opencontainers.image.description="Base build tools for VMAF fork dev-MCP container"
-LABEL org.opencontainers.image.source="https://github.com/lusoris/vmaf"
+LABEL org.opencontainers.image.source="https://github.com/VMAFx/vmafx"
 LABEL org.opencontainers.image.licenses="BSD-3-Clause-Plus-Patent"
 
 ARG DEBIAN_FRONTEND=noninteractive

docker/Dockerfile.production

@@ -29,7 +29,7 @@ ARG DEBIAN_FRONTEND=noninteractive
 
 LABEL org.opencontainers.image.title="vmafx-builder"
 LABEL org.opencontainers.image.description="Build stage: libvmaf + vmaf CLI (release/stripped)"
-LABEL org.opencontainers.image.source="https://github.com/lusoris/vmaf"
+LABEL org.opencontainers.image.source="https://github.com/VMAFx/vmafx"
 LABEL org.opencontainers.image.licenses="BSD-3-Clause-Plus-Patent"
 
 # Install only build-time deps; nothing carried into the runtime stage
@@ -102,7 +102,7 @@ FROM gcr.io/distroless/cc-debian12@sha256:aa0b7af67fa8211751ea6e00baa8373ba56cc1
 # placeholders here so local builds still produce labelled images.
 LABEL org.opencontainers.image.title="vmafx"
 LABEL org.opencontainers.image.description="VMAFX production runtime — libvmaf + vmaf CLI / MCP server"
-LABEL org.opencontainers.image.source="https://github.com/lusoris/vmaf"
+LABEL org.opencontainers.image.source="https://github.com/VMAFx/vmafx"
 LABEL org.opencontainers.image.licenses="BSD-3-Clause-Plus-Patent"
 LABEL org.opencontainers.image.vendor="Lusoris"
 

docker/Dockerfile.production-gpu

@@ -11,11 +11,11 @@
 #   vulkan     — adds Vulkan runtime ICD loader (libvulkan.so + lavapipe ICD JSON)
 #
 # Tag matrix (built by CI):
-#   ghcr.io/lusoris/vmafx:<tag>          (cpu default, same as Dockerfile.production)
-#   ghcr.io/lusoris/vmafx:<tag>-cuda12
-#   ghcr.io/lusoris/vmafx:<tag>-rocm6
-#   ghcr.io/lusoris/vmafx:<tag>-oneapi2026
-#   ghcr.io/lusoris/vmafx:<tag>-vulkan
+#   ghcr.io/vmafx/vmafx:<tag>          (cpu default, same as Dockerfile.production)
+#   ghcr.io/vmafx/vmafx:<tag>-cuda12
+#   ghcr.io/vmafx/vmafx:<tag>-rocm6
+#   ghcr.io/vmafx/vmafx:<tag>-oneapi2026
+#   ghcr.io/vmafx/vmafx:<tag>-vulkan
 #
 # ADR-0698: vmafx-production-dockerfile
 
@@ -164,7 +164,7 @@ FROM gcr.io/distroless/cc-debian12@sha256:aa0b7af67fa8211751ea6e00baa8373ba56cc1
 
 LABEL org.opencontainers.image.title="vmafx"
 LABEL org.opencontainers.image.description="VMAFX production GPU runtime"
-LABEL org.opencontainers.image.source="https://github.com/lusoris/vmaf"
+LABEL org.opencontainers.image.source="https://github.com/VMAFx/vmafx"
 LABEL org.opencontainers.image.licenses="BSD-3-Clause-Plus-Patent"
 LABEL org.opencontainers.image.vendor="Lusoris"
 

docs/adr/0002-merge-path-master-default.md

@@ -11,7 +11,7 @@ The fork carried multiple long-lived branches (`gpu-opt`, `sycl`, `master`) with
 
 ## Decision
 
-We will integrate via `gpu-opt → sycl → master`, make `master` the fork default on `origin` (`gh api -X PATCH repos/lusoris/vmaf -f default_branch=master`), and keep `upstream/master` fetchable via `git fetch upstream` for periodic syncs.
+We will integrate via `gpu-opt → sycl → master`, make `master` the fork default on `origin` (`gh api -X PATCH repos/VMAFx/vmafx -f default_branch=master`), and keep `upstream/master` fetchable via `git fetch upstream` for periodic syncs.
 
 ## Alternatives considered
 

docs/adr/0102-dnn-ep-selection-and-fp16-io.md

@@ -18,7 +18,7 @@ accepted at the public surface and then silently dropped by the backend:
   `ROCM` enum values existed but had no wiring at all.
 - `fp16_io` was a documented `bool` with no code path ever reading it.
 
-Issue [#30](https://github.com/lusoris/vmaf/issues/30) flagged this as
+Issue [#30](https://github.com/VMAFx/vmafx/issues/30) flagged this as
 an API correctness bug: callers that set `device = VMAF_DNN_DEVICE_CUDA`
 or `fp16_io = true` got a CPU fp32 session, with no log, no error, and
 no way to detect the silent downgrade. The same applies to
@@ -109,7 +109,7 @@ hosts without hardware fp16.
 
 ## References
 
-- Issue [#30 — VmafDnnDevice OPENVINO/ROCM + fp16_io accepted-but-ignored](https://github.com/lusoris/vmaf/issues/30)
+- Issue [#30 — VmafDnnDevice OPENVINO/ROCM + fp16_io accepted-but-ignored](https://github.com/VMAFx/vmafx/issues/30)
 - [ONNX Runtime C API — execution providers](https://onnxruntime.ai/docs/execution-providers/)
 - [ADR-0040](decisions-log.md) — multi-input DNN session API (the same `VmafDnnConfig` this ADR extends)
 - [ADR-0042 / ADR-0100](decisions-log.md) — docs-in-same-PR rules (this PR updates `docs/api/dnn.md` accordingly)