Inference Speed on Audio 10s (test)
42Inference Time (ms)LPA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LPAPulse Config=8/12, Precision=fp16, Backend=MLX, Hardware=Apple M4 Pro, Batch size=12026.03 | 42 | 1.07 | |
| Attention BaselinePrecision=fp16, Backend=PyTorch/MPS, Hardware=Apple M4 Pro, Batch size=12026.03 | 45 | — | |
| Attention BaselinePrecision=fp32, Backend=PyTorch/MPS, Hardware=Apple M4 Pro, Batch size=12026.03 | 47 | — | |
| LPAPulse Config=8/12, Precision=fp32, Backend=PyTorch/MPS, Hardware=Apple M4 Pro, Batch size=12026.03 | 59 | — | |
| LPAPulse Config=12/12, Precision=fp32, Backend=PyTorch/MPS, Hardware=Apple M4 Pro, Batch size=12026.03 | 65 | — |