Repository navigation

Tensor inference, opt-in vocal separation, GPU runtime, and cross-platform CI - #9

Merged

soficis merged 72 commits into

masterfrom

feat/tensor-inference

Oct 5, 2026

Merged

Tensor inference, opt-in vocal separation, GPU runtime, and cross-platform CI#9
soficis merged 72 commits into
masterfrom
feat/tensor-inference

Commits on Sep 30, 2026
- fix(renderers): search AUTOMIX_ASSET_ROOT before default PhaseLimiter roots
  
  soficis
  and
  claude
  committed
  cb9a41c
- feat(ai): tensor inference and opt-in BS-RoFormer vocal separation
  
  soficis
  and
  claude
  committed
  1b5610f
Commits on Oct 1, 2026
- fix(ui): convert licence string_views before building juce::String
  
  soficis
  and
  claude
  committed
  dcebb4d
- feat(ui): per-session Vocal Model toggle and separate dev-tools command
  
  soficis
  and
  claude
  committed
  dc2c919
- test(audioio): accept the initial meter value in the torn-read test
  
  soficis
  and
  claude
  committed
  97e2d09
- fix(ai): append CUDA/TensorRT through their V2 provider APIs
  
  soficis
  and
  claude
  committed
  ddc217b
- feat(ai): cancellable, GPU-accelerated tensor separation
  
  soficis
  and
  claude
  committed
  3e9d979
- feat(ai): gate fp32 GPU separation on device memory
  
  soficis
  and
  claude
  committed
  23a4f49
- feat(ai): on-demand per-user GPU runtime pack for CUDA
  
  soficis
  and
  claude
  committed
  05ef29d
- feat(ai): driver gate, model auto-upgrade and removal for the GPU runtime
  
  soficis
  and
  claude
  committed
  8a42f1b
- build: Windows release ZIP and lean PhaseLimiter asset staging
  
  soficis
  and
  claude
  committed
  1aa7f53
- fix(renderers): make PhaseLimiter work, opt-in, and leak-free
  
  soficis
  and
  claude
  committed
  60265fa
- feat(ai): per-user model hub with one-time migration
  
  soficis
  and
  claude
  committed
  8682412
- fix(ai): unstick stale model registry entries; correct quantized-CUDA claim
  
  soficis
  and
  claude
  committed
  1d09f7c
Commits on Oct 2, 2026
- fix(ui): rescan model packs after the in-app GPU vocal-model upgrade
  
  soficis
  and
  claude
  committed
  dc4882d
- feat(ai): unify session configuration and fix CoreML ANE options
  soficis
  committed
  05f6742
- feat(ai): add webgpu pure logic, update priority chain order, and plumb gpu allow-lists
  soficis
  committed
  22935db
- fix(ai): untuned tensor sessions, BS-RoFormer allow-list default, hardened OrtRuntime and probe cache
  
  soficis
  and
  Muse Spark 1.3
  committed
  613d820
Commits on Oct 3, 2026
Commits on Oct 4, 2026
Commits on Oct 5, 2026

Commits

Commits on Sep 30, 2026

Commits on Oct 1, 2026

Commits on Oct 2, 2026

Commits on Oct 3, 2026

Commits on Oct 4, 2026

Commits on Oct 5, 2026