-
Switch to mlprogram backend with ANE/GPU support
Clément Duhamel
Changes:
- Converted both encoders to CoreML mlprogram (modern backend)
- Disabled fuse_transpose_matmul pass to fix NaN on transformer attention
- Forced FLOAT32 precision to avoid convolution overflow
- Target: macOS 14+ / iOS 17+ for full ANE compatibility
- Numerically verified: audio cos_sim = 1.00000036, text cos_sim = 1.00000012
- Updated model card (README.md) and Git LFS tracking for mlprogram -
Switch to mlprogram backend with ANE/GPU support
Clément Duhamel
Changes:
- Converted both encoders to CoreML mlprogram (modern backend)
- Disabled fuse_transpose_matmul pass to fix NaN on transformer attention
- Forced FLOAT32 precision to avoid convolution overflow
- Target: macOS 14+ / iOS 17+ for full ANE compatibility
- Numerically verified: audio cos_sim = 1.00000036, text cos_sim = 1.00000012
Note: .mlmodel (neuralnetwork) replaced by .mlpackage (mlprogram)
Git LFS tracking updated for weight.bin files -
Upload TinyCLAP CoreML models
Velvox Builder
Audio encoder: TinyCLAP_AudioEncoder.mlmodel (~17 MB)
Text encoder: TinyCLAP_TextEncoder.mlmodel (~422 MB)
Preprocessing: mel_filter_bank.json
