Skip to content

Add full beat_this! ai model to mobile. - #132

Merged
SFG5453 merged 5 commits into
betafrom
feat/full-mobile-ai-model
Sep 10, 2026
Merged

SFG5453 merged 5 commits into
betafrom
feat/full-mobile-ai-model

Conversation

@SFG5453

@SFG5453 SFG5453 commented Sep 10, 2026

Copy link
Copy Markdown
Owner

No description provided.

Package the QNN-enabled ONNX Runtime and matching HTP libraries, and declare access to the vendor FastRPC transport. Probe NPU availability with validated inference and CPU fallback disabled, expose the result through the app graph, and add unit and device coverage.

Implemented-with: Codex
Replace the shipping beat model with final0 dynamic INT8, pad fixed-size inference windows, and refresh the extracted model cache. Remove QNN runtime packaging and NPU detection, and use stock ONNX Runtime with explicit CPU execution.

Add quantization benchmarks, accuracy evaluation tools, and experiment documentation. The NPU attempt failed to deliver acceptable production accuracy despite successful cached inference, so production remains on CPU.

Fixes #126
@SFG5453 SFG5453 self-assigned this Sep 10, 2026
@SFG5453
SFG5453 merged commit 677d766 into beta Sep 10, 2026
3 checks passed
@SFG5453
SFG5453 deleted the feat/full-mobile-ai-model branch September 10, 2026 05:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant