QVAC-23075 feat[api]: accept image_no_upscale in the addon load config by yingying0906 · Pull Request #3725 · tetherto/qvac

yingying0906 · 2026-08-07T16:39:15Z

🎯 What problem does this PR solve?

The VisionPsy Flash checkpoint and the base one are indistinguishable to the addon. Their mmprojs declare identical vision hparams, so a Flash checkpoint loaded without this key silently runs base preprocessing.
That changes the image token count, so it moves both accuracy and encode time.

📝 How does it solve it?

image_no_upscale on the addon load config, forwarded to the vision context, so a caller can override the idefics3-style preprocessing rule the GGUF declares instead of being stuck with it.
LoadConfigHandlers.cpp parses 0/off/false or 1/on/true into common_params, and rejects anything else with InvalidArgument.
MtmdLlmContext::initVisionContext copies it into mtmd_context_params, next to image_tile_mode.
Leaving the key out keeps fabric's -1 model default, so existing callers are unaffected.
Needs QVAC-23075 feat: add VisionPsy Nano support and its Flash preprocessing rule qvac-fabric-llm.cpp#205, which adds the field to common_params and mtmd_context_params. That has now merged and been tagged as v10069.1.0 — see the rollout section below for the current build state.

🧪 How was it tested?

test_load_config_handlers.cpp covers the parse alongside the other load-config cases.

🔌 API Changes

// llm-llamacpp load config, new optional key
{
  'image-no-upscale': 'on' // or image_no_upscale; 0/off/false or 1/on/true
}

🚚 qvac-fabric 10069.1.0 rollout rides this PR

Phase B of the fabric rollout landed here rather than on a separate bump PR, so the consumer code and
the fabric version it needs meet on one PR.


Fabric PR	tetherto/qvac-fabric-llm.cpp#205 — VisionPsy Nano support + Flash preprocessing rule
Tag	`v10069.1.0` at `50713f4dfc54b1fe09826a343527e83d90f5df62` (head of `temp-10069`)
Registry PR	tetherto/qvac-registry-vcpkg#317 — must merge first

Three commits were added on top of the feature work:

Overlay removed — the temporary Phase A qvac-fabric overlay port and its overlay-ports keys
are reverted; the 7 consumers resolve fabric from the registry again.
main merged in — the branch predated the @qvac/llm-llamacpp 0.43.0 release, so the version
bumps had to be applied on top of current main. Clean merge, no conflicts.
Fabric bump across all 7 consumers — version>= 10069.0.0 → 10069.1.0, plus package
versions and changelog entries:

Package	Version
`@qvac/embed-llamacpp`	0.32.0 → 0.33.0
`@qvac/fabric`	0.4.0 → 0.5.0
`@qvac/llm-llamacpp`	0.43.0 → 0.44.0
`@qvac/model-fit`	0.1.0 → 0.2.0
`@qvac/ocr-ggml`	0.16.0 → 0.17.0
`@qvac/translation-nmtcpp`	0.8.0 → 0.9.0
`@qvac/vla-ggml`	0.19.0 → 0.20.0

No default-registry.baseline changes anywhere.

⚠️ What reviewers should know

Registry PR (chore: bump qvac-fabric to 10069.1.0 qvac-registry-vcpkg#317) is now merged and qvac-fabric 10069.1.0 is published, so version>=: 10069.1.0
resolves and CI has been retriggered. cpp and desktop are green. test-darwin-x64 fails on
Qwen3 sliding context hard-fails stale reasoning compaction in reasoning.test.js, a known
darwin-x64 issue on main that this PR does not touch and that also reproduces on
tmp-ext-rebase-fabric-b10297.
Mobile integration tests: https://github.com/tetherto/qvac/actions/runs/32117504412
Dispatched through the CI router at the branch ref, so the phones run this PR's own prebuilds
rather than the published package. Android passed. On iOS every shard passed except lightB on
iPhone 16, where the Device Farm host could not reach PyPI and failed to install its own tooling
before any test ran. The same shard passed on iPhone 17, and the iOS leg has been rerun.
Merging this PR now also publishes 7 packages. It is a release PR as well as a feature PR.
@qvac/llm-llamacpp 0.44.0 is what ships image_no_upscale, and its changelog entry documents both
that option and the fabric bump.

To see the specific tasks where the Asana app for GitHub is being used, see below:
- https://app.asana.com/0/0/1217241821002485

github-actions · 2026-08-07T16:39:45Z

License compliance — clean

No new dependency license findings in this PR.

Warn-only (shadow) mode — this check does not block merges yet.

Updated automatically by the canonical license compliance workflow.

NOTICE presence (advisory)

Missing NOTICE (advisory, does not block):

./.github/actions/release-merge-guard
./docs/website
./packages/ggml-coload-smoke
./packages/fabric/test/integration
./packages/inference-addon-cpp/mobile
./packages/sdk/e2e
./packages/llm-llamacpp/benchmarks/performance
./packages/llm-llamacpp/benchmarks/server
./packages/vla-ggml/sim/server
./packages/embed-llamacpp/benchmarks/performance
./packages/embed-llamacpp/benchmarks/server
./packages/asr-ggml/benchmarks/server

github-actions · 2026-08-07T16:39:57Z

Workflow security (shadow mode)

zizmor found 1108 finding(s) in .github/ (highest severity: high). This check is warn-only and does not block the merge.

Findings are annotated inline on the changed files and listed in the job summary.

Reproduce locally:

pipx run zizmor==1.27.0 --offline .github/

yingying0906 · 2026-08-07T16:41:14Z

Depends on tetherto/qvac-fabric-llm.cpp#205 and the registry publish that follows it. CI here stays red until then, by design. Build and device evidence is on bench/QVAC-23075-visionpsy-vlm, which carries the vcpkg overlay. It was PR #3726, now closed, and those run records are still readable there.

github-actions · 2026-08-12T13:16:46Z

Review Status

Current Status: ✅ APPROVED
Approvals so far: Team Lead: 1, Member: 1

tobi-legan · 2026-08-12T15:23:58Z

Review follow-up: remaining benchmark hardening + QA gaps

I see the cliArgs allowlist bypass is already covered inline, so not duplicating it here. One remaining benchmark hardening item: the workflow prints cli-model.env, which includes LLM_URL and MMPROJ_URL. For arbitrary json: URL/S3 sources, those may be presigned or otherwise private direct links, so please redact URL variables from the log or print only model names/provenance before sourcing the env file.

QA gaps I would like covered before treating the benchmark plumbing as review-ready:

Add unit coverage across models.cjs, resolve-cli-model.cjs, and cli-fixture-runner.cjs for cliArgs token preservation, including whitespace-bearing values and --flag=value --forbidden bypass attempts.
Add a benchmark-script smoke test for several-sources with matrix_sources=fabric@... and non-default matrix_models, verifying the CLI-only path downloads the requested blobs and passes model-specific args only to fabric.
Add parser fixture tests for stdout-parser.js and aggregate.js using real llama-mtmd-cli log snippets, including batch timing, chunk counts, and mixed helper/batch lines to prevent double-counting regressions.
Add a contract generation check that asserts image_no_upscale appears consistently in the TypeScript schema, JSON contract, and generated Python models.

tobi-legan

Requesting changes based on the benchmark plumbing issues and QA gaps already posted.

Blocking item: the cliArgs validation can be bypassed through the join/split handoff, allowing benchmark-fixed CLI options to be overridden in the fabric CLI leg. Please preserve arg boundaries or revalidate after splitting.

Also please address the URL logging hardening and add focused coverage for models.cjs, resolve-cli-model.cjs, cli-fixture-runner.cjs, CLI-only several-sources, and the stdout/aggregate parser fixtures before this is merge-ready.

yingying0906 · 2026-08-13T02:38:04Z

@dev-nid @tobi-legan

The cliArgs bypass is fixed in ddc0e63, one element can no longer carry whitespace. The provenance issue is fixed in 3c607e6. Fabric side has the review commits on tetherto/qvac-fabric-llm.cpp#205, ending at 295226e2d. Ready for another look when you are. The SHAs here are post rebase onto main.

github-actions · 2026-08-17T00:30:40Z

🧪 C++ Test Coverage Report

Coverage:

📊 Detailed Coverage

Filename                         Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
NmtLazyInitializeBackend.cpp          99                20    79.80%          11                 1    90.91%         157                36    77.07%          66                21    68.18%
NmtLazyInitializeBackend.hpp           2                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
TranslationModel.cpp                 296               168    43.24%          28                 8    71.43%         506               213    57.91%         181               122    32.60%
TranslationModel.hpp                   1                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
nmt.cpp                               72                22    69.44%           9                 1    88.89%         137                28    79.56%          44                16    63.64%
nmt.hpp                               51                 4    92.16%          11                 2    81.82%          53                 4    92.45%          28                 0   100.00%
nmt_beam_search.cpp                  116                25    78.45%          10                 3    70.00%         254                32    87.40%          76                19    75.00%
nmt_graph_decoder.cpp                164                78    52.44%          15                 7    53.33%         540               161    70.19%         112                69    38.39%
nmt_graph_encoder.cpp                 54                13    75.93%           3                 0   100.00%         268                33    87.69%          37                16    56.76%
nmt_loader.cpp                       270                67    75.19%          14                 0   100.00%         774                97    87.47%         161                67    58.39%
nmt_state_backend.cpp                253                94    62.85%          21                 0   100.00%         489               128    73.82%         165                87    47.27%
nmt_tokenization.cpp                  88                21    76.14%           8                 0   100.00%         135                36    73.33%          61                26    57.38%
nmt_utils.cpp                        120                89    25.83%           8                 3    62.50%         180               134    25.56%          78                63    19.23%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                               1586               601    62.11%         140                25    82.14%        3495               902    74.19%        1009               506    49.85%

github-actions · 2026-08-17T10:21:12Z

🧪 C++ Test Coverage Report

Coverage:

📊 Detailed Coverage

Filename                         Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
NmtLazyInitializeBackend.cpp          99                20    79.80%          11                 1    90.91%         157                36    77.07%          66                21    68.18%
NmtLazyInitializeBackend.hpp           2                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
TranslationModel.cpp                 296               168    43.24%          28                 8    71.43%         506               213    57.91%         181               122    32.60%
TranslationModel.hpp                   1                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
nmt.cpp                               72                22    69.44%           9                 1    88.89%         137                28    79.56%          44                16    63.64%
nmt.hpp                               51                 4    92.16%          11                 2    81.82%          53                 4    92.45%          28                 0   100.00%
nmt_beam_search.cpp                  116                25    78.45%          10                 3    70.00%         254                32    87.40%          76                19    75.00%
nmt_graph_decoder.cpp                164                78    52.44%          15                 7    53.33%         540               161    70.19%         112                69    38.39%
nmt_graph_encoder.cpp                 54                13    75.93%           3                 0   100.00%         268                33    87.69%          37                16    56.76%
nmt_loader.cpp                       270                67    75.19%          14                 0   100.00%         774                97    87.47%         161                67    58.39%
nmt_state_backend.cpp                253                94    62.85%          21                 0   100.00%         489               128    73.82%         165                87    47.27%
nmt_tokenization.cpp                  88                21    76.14%           8                 0   100.00%         135                36    73.33%          61                26    57.38%
nmt_utils.cpp                        120                89    25.83%           8                 3    62.50%         180               134    25.56%          78                63    19.23%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                               1586               601    62.11%         140                25    82.14%        3495               902    74.19%        1009               506    49.85%

VisionPsy answers count-fingers wrong. This does not fix that. It asks a wording the model gets right instead of asserting on the wording it fails, which is a workaround and worth calling one. It is defensible here because the test is "continuous batching MTMD: mixed image+text batch processes all slot types correctly". It covers batch admission and slot scheduling. One text-only slot returning a wrong answer says nothing about either, so gating this test on the answer to a general-knowledge question was testing the wrong thing. It is not evidence the model is fine, and the failure is real. Why a separate vlmUser instead of editing user. CASES feeds two paths with different system prompts: buildPrompt() to Llama-3.2-1B with the verbose 64-word instruction, buildVlmBatchItem() to the VLM pair with the one-word instruction. The two have no wording in common, so editing user just moves the failure. Measured with llama-cli at the same greedy settings the tests use, holding the frame at "How many fingers are on X? Answer with one word.": one typical human hand metamorphs a human hand 5 one human hand metamorphs the human hand 5 a typical human hand metamorphs an adult human hand 5 one hand metamorphs a normal human hand 5 a single human hand 5 your human hand 5 "one" and "typical" break VisionPsy and are exactly what Llama-3.2-1B needs, which answers "Ten" with them and "Fifty" without. SmolVLM2 answers 10 either way. The frame matters as much as the modifier: "are there on a single human hand" returns "No fingers" where "are on a single human hand" returns 5, so re-measure against both models before editing either wording. Reproduced on the research team's own llama.cpp build, upstream 08023072e plus visionpsy-nano.diff, identical on all 18 cells of a 6-wording by 3-model sweep, so this is the model and not the port. Reported separately to research. Verified locally against a fresh addon build, both pairs green: continuous-batching.test.js 13/13 tests and 167/167 asserts for the default visionpsy pair and for QVAC_VLM_MODEL=smolvlm2.

github-actions · 2026-08-17T12:42:22Z

Mobile integration tests — @qvac/model-fit (Android)

Result: passed

metric	value
Devices passed	3
Devices failed	0
Test cases total	9
Test cases passed	9
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:46:57Z

Mobile integration tests — @qvac/translation-nmtcpp (Android)

Result: passed

metric	value
Devices passed	2
Devices failed	0
Test cases total	6
Test cases passed	6
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:50:44Z

Mobile integration tests — @qvac/vla-ggml (Android)

Result: passed

metric	value
Devices passed	9
Devices failed	0
Test cases total	27
Test cases passed	27
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:50:50Z

Mobile integration tests — @qvac/ocr-ggml (Android)

Result: passed

metric	value
Devices passed	9
Devices failed	0
Test cases total	27
Test cases passed	27
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:54:55Z

Mobile integration tests — @qvac/model-fit (iOS)

Result: passed

metric	value
Devices passed	2
Devices failed	0
Test cases total	6
Test cases passed	6
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:55:31Z

Mobile integration tests — @qvac/translation-nmtcpp (iOS)

Result: passed

metric	value
Devices passed	3
Devices failed	0
Test cases total	9
Test cases passed	9
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:57:57Z

Mobile integration tests — @qvac/ocr-ggml (iOS)

Result: passed

metric	value
Devices passed	6
Devices failed	0
Test cases total	18
Test cases passed	18
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:58:09Z

Mobile integration tests — @qvac/vla-ggml (iOS)

Result: passed

metric	value
Devices passed	6
Devices failed	0
Test cases total	18
Test cases passed	18
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T12:59:46Z

🧪 C++ Test Coverage Report

Coverage:

📊 Detailed Coverage

Filename                         Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
NmtLazyInitializeBackend.cpp          99                20    79.80%          11                 1    90.91%         157                36    77.07%          66                21    68.18%
NmtLazyInitializeBackend.hpp           2                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
TranslationModel.cpp                 296               168    43.24%          28                 8    71.43%         506               213    57.91%         181               122    32.60%
TranslationModel.hpp                   1                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
nmt.cpp                               72                22    69.44%           9                 1    88.89%         137                28    79.56%          44                16    63.64%
nmt.hpp                               51                 4    92.16%          11                 2    81.82%          53                 4    92.45%          28                 0   100.00%
nmt_beam_search.cpp                  116                25    78.45%          10                 3    70.00%         254                32    87.40%          76                19    75.00%
nmt_graph_decoder.cpp                164                78    52.44%          15                 7    53.33%         540               161    70.19%         112                69    38.39%
nmt_graph_encoder.cpp                 54                13    75.93%           3                 0   100.00%         268                33    87.69%          37                16    56.76%
nmt_loader.cpp                       270                67    75.19%          14                 0   100.00%         774                97    87.47%         161                67    58.39%
nmt_state_backend.cpp                253                94    62.85%          21                 0   100.00%         489               128    73.82%         165                87    47.27%
nmt_tokenization.cpp                  88                21    76.14%           8                 0   100.00%         135                36    73.33%          61                26    57.38%
nmt_utils.cpp                        120                89    25.83%           8                 3    62.50%         180               134    25.56%          78                63    19.23%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                               1586               601    62.11%         140                25    82.14%        3495               902    74.19%        1009               506    49.85%

github-actions · 2026-08-17T13:08:17Z

Mobile integration tests — @qvac/embed-llamacpp (Android)

Result: passed

metric	value
Devices passed	6
Devices failed	0
Test cases total	18
Test cases passed	18
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T13:14:39Z

Mobile integration tests — @qvac/embed-llamacpp (iOS)

Result: passed

metric	value
Devices passed	4
Devices failed	0
Test cases total	12
Test cases passed	12
Test cases failed	0
Test cases skipped	0

View workflow run

github-actions · 2026-08-17T13:22:30Z

Mobile integration tests — @qvac/llm-llamacpp (iOS)

Result: failed

metric	value
Devices passed	25
Devices failed	1
Test cases total	78
Test cases passed	77
Test cases failed	1
Test cases skipped	0

View workflow run

….1.0 is published Reverts the two overlay-validation commits (7210e53 and its re-pin 40a1576). qvac-fabric v10069.1.0 is tagged and published to the registry, so the 7 fabric consumers resolve it from the registry again instead of the local overlay portfile. Note: until the bundled consumer bump lands, the version>= floors here still read 10069.0.0, so this branch builds against the previously published fabric.

github-actions · 2026-08-17T18:31:55Z

Security baseline — findings detected

TruffleHog: success
CodeQL: failure
Severity threshold: high

See the job summary and the repository Security tab for details.

This comment is updated automatically by the canonical security workflow.

Brings the branch up to date with origin/main (41 commits) ahead of the qvac-fabric 10069.1.0 dependency bump. Clean merge, no conflicts. Needed because the branch predated the @qvac/llm-llamacpp 0.43.0 release: the fabric bump has to be applied on top of the current versions, or the bumped numbers collide with main and the 0.43.0 changelog entry goes missing.

Bumps the qvac-fabric vcpkg dependency floor from 10069.0.0 to 10069.1.0 for all 7 fabric consumers, with the matching package version bumps and changelog entries. qvac-fabric 10069.1.0 adds VisionPsy Nano support and its Flash preprocessing rule (tetherto/qvac-fabric-llm.cpp#205) — the fabric side this PR's image_no_upscale load option depends on. - embed-llamacpp 0.32.0 -> 0.33.0 - fabric 0.4.0 -> 0.5.0 - llm-llamacpp 0.43.0 -> 0.44.0 - model-fit 0.1.0 -> 0.2.0 - ocr-ggml 0.16.0 -> 0.17.0 - translation-nmtcpp 0.8.0 -> 0.9.0 - vla-ggml 0.19.0 -> 0.20.0 llm-llamacpp's entry also documents image_no_upscale, since this bump is what creates the 0.44.0 release that publishes it. The other 6 are fabric-only with no API change. Registry publish: tetherto/qvac-registry-vcpkg#317. CI cannot resolve version>= 10069.1.0 until that merges.

github-actions · 2026-08-17T21:22:26Z

🧪 C++ Test Coverage Report

Coverage:

📊 Detailed Coverage

Filename                         Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
NmtLazyInitializeBackend.cpp          99                20    79.80%          11                 1    90.91%         157                36    77.07%          66                21    68.18%
NmtLazyInitializeBackend.hpp           2                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
TranslationModel.cpp                 296               168    43.24%          28                 8    71.43%         506               213    57.91%         181               122    32.60%
TranslationModel.hpp                   1                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
nmt.cpp                               72                22    69.44%           9                 1    88.89%         137                28    79.56%          44                16    63.64%
nmt.hpp                               51                 4    92.16%          11                 2    81.82%          53                 4    92.45%          28                 0   100.00%
nmt_beam_search.cpp                  116                25    78.45%          10                 3    70.00%         254                32    87.40%          76                19    75.00%
nmt_graph_decoder.cpp                164                78    52.44%          15                 7    53.33%         540               161    70.19%         112                69    38.39%
nmt_graph_encoder.cpp                 54                13    75.93%           3                 0   100.00%         268                33    87.69%          37                16    56.76%
nmt_loader.cpp                       270                67    75.19%          14                 0   100.00%         774                97    87.47%         161                67    58.39%
nmt_state_backend.cpp                253                94    62.85%          21                 0   100.00%         489               128    73.82%         165                87    47.27%
nmt_tokenization.cpp                  88                21    76.14%           8                 0   100.00%         135                36    73.33%          61                26    57.38%
nmt_utils.cpp                        120                89    25.83%           8                 3    62.50%         180               134    25.56%          78                63    19.23%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                               1586               601    62.11%         140                25    82.14%        3495               902    74.19%        1009               506    49.85%

github-actions · 2026-08-18T10:10:53Z

🧪 C++ Test Coverage Report

Coverage:

📊 Detailed Coverage

Filename                         Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
NmtLazyInitializeBackend.cpp          99                20    79.80%          11                 1    90.91%         157                36    77.07%          66                21    68.18%
NmtLazyInitializeBackend.hpp           2                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
TranslationModel.cpp                 296               168    43.24%          28                 8    71.43%         506               213    57.91%         181               122    32.60%
TranslationModel.hpp                   1                 0   100.00%           1                 0   100.00%           1                 0   100.00%           0                 0         -
nmt.cpp                               72                22    69.44%           9                 1    88.89%         137                28    79.56%          44                16    63.64%
nmt.hpp                               51                 4    92.16%          11                 2    81.82%          53                 4    92.45%          28                 0   100.00%
nmt_beam_search.cpp                  116                25    78.45%          10                 3    70.00%         254                32    87.40%          76                19    75.00%
nmt_graph_decoder.cpp                164                78    52.44%          15                 7    53.33%         540               161    70.19%         112                69    38.39%
nmt_graph_encoder.cpp                 54                13    75.93%           3                 0   100.00%         268                33    87.69%          37                16    56.76%
nmt_loader.cpp                       270                67    75.19%          14                 0   100.00%         774                97    87.47%         161                67    58.39%
nmt_state_backend.cpp                253                94    62.85%          21                 0   100.00%         489               128    73.82%         165                87    47.27%
nmt_tokenization.cpp                  88                21    76.14%           8                 0   100.00%         135                36    73.33%          61                26    57.38%
nmt_utils.cpp                        120                89    25.83%           8                 3    62.50%         180               134    25.56%          78                63    19.23%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                               1586               601    62.11%         140                25    82.14%        3495               902    74.19%        1009               506    49.85%

iancris · 2026-08-18T10:57:00Z

/review

This was referenced Aug 7, 2026

QVAC-23075 test[skiplog]: DO NOT MERGE, overlay-pinned CI proof for VisionPsy Nano #3726

Closed

QVAC-23075 feat: add VisionPsy Nano support and its Flash preprocessing rule tetherto/qvac-fabric-llm.cpp#205

Merged

yingying0906 force-pushed the feat/QVAC-23075-visionpsy branch 7 times, most recently from 7d4b2ce to 2cc8d8e Compare August 12, 2026 03:57

yingying0906 marked this pull request as ready for review August 12, 2026 13:16

yingying0906 requested review from a team as code owners August 12, 2026 13:16

yingying0906 had a problem deploying to release August 12, 2026 13:17 — with GitHub Actions Failure

dev-nid reviewed Aug 12, 2026

View reviewed changes

Comment thread packages/llm-llamacpp/benchmarks/vlm-benchmark/models.cjs Outdated

dev-nid reviewed Aug 12, 2026

View reviewed changes

Comment thread packages/llm-llamacpp/benchmarks/vlm-benchmark/harness.cjs Outdated

tobi-legan requested changes Aug 12, 2026

View reviewed changes

yingying0906 temporarily deployed to release August 12, 2026 15:36 — with GitHub Actions Inactive

yingying0906 had a problem deploying to release August 12, 2026 15:36 — with GitHub Actions Failure

yingying0906 had a problem deploying to release August 12, 2026 15:44 — with GitHub Actions Failure

yingying0906 temporarily deployed to release August 12, 2026 15:44 — with GitHub Actions Inactive

yingying0906 had a problem deploying to release August 12, 2026 15:57 — with GitHub Actions Failure

yingying0906 temporarily deployed to release August 12, 2026 15:57 — with GitHub Actions Inactive

yingying0906 had a problem deploying to release August 13, 2026 02:48 — with GitHub Actions Failure

yingying0906 temporarily deployed to release August 13, 2026 02:49 — with GitHub Actions Inactive

iancris mentioned this pull request Aug 17, 2026

QVAC-23075 feat[api]: bump qvac-fabric to 10069.1.0 across consumers #3907

Closed

iancris added 2 commits August 17, 2026 19:11

iancris mentioned this pull request Aug 17, 2026

chore: bump qvac-fabric to 10069.1.0 tetherto/qvac-registry-vcpkg#317

Merged

iancris approved these changes Aug 18, 2026

View reviewed changes

Merge branch 'main' into feat/QVAC-23075-visionpsy

40ccafa

gianni-cor approved these changes Aug 18, 2026

View reviewed changes

Merge branch 'main' into feat/QVAC-23075-visionpsy

af0ff7d

iancris mentioned this pull request Aug 18, 2026

QVAC-22630 chore: bump model-fit to 0.2.1 and record the fit-context fix #3917

Merged

yingying0906 mentioned this pull request Aug 20, 2026

QVAC-23811 doc: add VisionPsy Nano and image_no_upscale to multimodal page #3959

Merged

Uh oh!

Conversation

yingying0906 commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🎯 What problem does this PR solve?

📝 How does it solve it?

🧪 How was it tested?

🔌 API Changes

🚚 qvac-fabric 10069.1.0 rollout rides this PR

⚠️ What reviewers should know

Uh oh!

github-actions Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

License compliance — clean

NOTICE presence (advisory)

Uh oh!

github-actions Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Workflow security (shadow mode)

Uh oh!

yingying0906 commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Aug 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review Status

Uh oh!

Uh oh!

Uh oh!

tobi-legan commented Aug 12, 2026

Review follow-up: remaining benchmark hardening + QA gaps

Uh oh!

tobi-legan left a comment

Choose a reason for hiding this comment

Uh oh!

yingying0906 commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Aug 17, 2026

🧪 C++ Test Coverage Report

Uh oh!

github-actions Bot commented Aug 17, 2026

🧪 C++ Test Coverage Report

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/model-fit (Android)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/translation-nmtcpp (Android)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/vla-ggml (Android)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/ocr-ggml (Android)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/model-fit (iOS)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/translation-nmtcpp (iOS)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/ocr-ggml (iOS)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/vla-ggml (iOS)

Uh oh!

github-actions Bot commented Aug 17, 2026

🧪 C++ Test Coverage Report

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/embed-llamacpp (Android)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/embed-llamacpp (iOS)

Uh oh!

github-actions Bot commented Aug 17, 2026

Mobile integration tests — @qvac/llm-llamacpp (iOS)

Uh oh!

yingying0906 commented Aug 7, 2026 •

edited

Loading

github-actions Bot commented Aug 7, 2026 •

edited

Loading

github-actions Bot commented Aug 7, 2026 •

edited

Loading

yingying0906 commented Aug 7, 2026 •

edited

Loading

github-actions Bot commented Aug 12, 2026 •

edited

Loading

yingying0906 commented Aug 13, 2026 •

edited

Loading