QVAC-17129 feat: add dynamic backend loading to nmtcpp GGML backend for GPU acceleration#1617
Merged
iancris merged 9 commits intoApr 22, 2026
Merged
Conversation
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
1 similar comment
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Bare runtime does not have Node.js built-in 'path' module. All other files in the package use 'bare-path'. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…lationModel Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…extraction Stop erasing backendsdir/openclcachedir from config_ in TranslationModel::load so reload() re-initializes the backend with the original path instead of falling back to the default loader. Without this, GPU backends disappeared silently after any model reload on Android. Use std::get_if<std::string> for backendsdir/openclcachedir extraction in TranslationModel and PivotTranslationModel; on type mismatch log a warning and fall through instead of propagating bad_variant_access out of the load path. Default backendsDir in JS only when the field is missing (=== undefined) rather than on any falsy value, so callers passing "" to force the default system backend loader are respected. Refs QVAC-17129.
PivotTranslationModel only uses Bergamot — GGML dynamic backend loading is for the IndicTrans (GGML) path, which is a different inference engine. The earlier backendsHandle_ initialization in the pivot constructor was initializing GGML backends even when none were used. Remove the backend config extraction block, the backendsHandle_ member and the NmtLazyInitializeBackend include. Sub-model TranslationModel instances still handle their own backend init on load() for the GGML path. Refs QVAC-17129.
Contributor
✅ E2E Mobile Test Results - iOSOverall Status: PASSED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
Contributor
✅ E2E Mobile Test Results - AndroidOverall Status: PASSED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
olyasir
approved these changes
Apr 17, 2026
Alok-Ranjan23
approved these changes
Apr 20, 2026
5 tasks
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
Contributor
✅ E2E Mobile Test Results - iOSOverall Status: PASSED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
✅ E2E Mobile Test Results - AndroidOverall Status: PASSED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
Author
|
/review |
Contributor
❌ E2E Mobile Test Results - iOSOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - AndroidOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - iOSOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - AndroidOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - iOSOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - AndroidOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - iOSOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
Contributor
❌ E2E Mobile Test Results - AndroidOverall Status: FAILED Test Summary
Links
Automated E2E mobile testing powered by AWS Device Farm |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Changes
New files
addon/src/model-interface/NmtLazyInitializeBackend.hpp— Lazy-init singleton with RAII NmtBackendsHandleaddon/src/model-interface/NmtLazyInitializeBackend.cpp— Implementation: ggml_backend_load_all_from_path(), OpenCL cache dir, ref-counted lifecycleaddon/tests/test_nmt_lazy_init_backend.cpp— 15 GoogleTest unit testsModified files
TranslationModel.hpp/cpp— Extract backendsdir/openclcachedir from config in load(), init handle before nmt_init_from_file_with_params()PivotTranslationModel.hpp/cpp— Init backends handle in constructor before sub-model load()index.js— Default backendsDir to prebuilds pathCMakeLists.txt— Add BACKENDS_SUBDIR, BACKEND_DL_LIBS loop, new source fileTest plan
CI: https://github.com/tetherto/qvac/actions/runs/24501619382
Asana
https://app.asana.com/0/1212638335655990/1214055988275114