| { |
| "model_name": "LavaSR-v2", |
| "target_hardware": "AX650", |
| "npu_mode": "NPU3", |
| "compiler": "Pulsar2 8.0", |
| "denoiser": { |
| "encoder": { |
| "file": "den_encoder.axmodel", |
| "input": {"name": "erb_feat", "shape": [1, 1, 301, 129]}, |
| "outputs": ["feat", "en0", "en1", "en2", "en3", "en4"] |
| }, |
| "decoder": { |
| "file": "den_decoder.axmodel", |
| "inputs": ["feat", "en0", "en1", "en2", "en3", "en4"], |
| "output": {"name": "mask", "shape": [1, 1, 301, 129]} |
| }, |
| "chunk_seconds": 4.8, |
| "chunk_samples": 76800, |
| "overlap": 0.25, |
| "host": "STFT/ERB/mask/ISTFT + DPGRNN (numpy)" |
| }, |
| "enhancer": { |
| "file": "enhancer.axmodel", |
| "input": {"name": "mel", "shape": [1, 80, 450]}, |
| "output": {"name": "head_out", "shape": [1, 2050, 450]}, |
| "chunk_seconds": 4.8, |
| "chunk_samples": 230400, |
| "overlap": 0.5, |
| "host": "44.1 kHz mel config on 48 kHz samples, exp/cos/sin/ISTFT + LR merge" |
| } |
| } |
|
|
