HTDemucs (Core ML, FP32), unpacked .mlpackage
HTDemucs 4-stem source separation (drums, bass, other, vocals) as a Core ML .mlpackage, FP32.
Converted by slaughters85j/htdemucs-coreml
(MIT); the original model is facebookresearch/demucs (MIT).
This repo holds the package unpacked so the files can be downloaded one by one, with resume, by the iPad app Minus One. The files are unchanged.
Interface: input mix [1, 2, 343980] (stereo, 44.1 kHz), output sources [1, 8, 343980].
| File | Bytes | SHA-256 |
|---|---|---|
Manifest.json |
617 | 5454ab98d4780d17a1b9d1570c573b96378f06e2e6ea5f8ee6c203dfabececbc |
Data/com.apple.CoreML/model.mlmodel |
1 476 469 | b6a93ae7836e44747392ef703f7ffb586c594f1d6bcfe6702d0e796e702faa75 |
Data/com.apple.CoreML/weights/weight.bin |
423 198 848 | c9380b228eda13ca18f13bcde83fb38b2aa6d03cae7c5454613b96a1461bb42e |
- Downloads last month
- 23
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support