Detailed model settings for Diffwave
| Learning_rate: 2e-4 | ||
|---|---|---|
| Data length L: 2 × 16k | ||
| Input (B, 1, L) | Diffusion-step embedding (B, 128) | |
| After Conv1x1 (B, 128, L) | After FC: (B, 512) | |
| After FC: (B, 512) | ||
| Residual block (30 layers) | After FC: (B, 128) | |
| Broadcast: (B, 128, L) | ||
| Dilated Conv setting: kernel size: 3, dilation: 2c mod 10 | ||
| After Dilated Conv: (B, 256, L) | ||
| After left Conv1x1: (B, 128, L) | After right Conv1x1: (B, 256, L) | |
| Output | After Conv1x1 (B, 256, L) | |
| After Conv1x1 (B, 1, L) | ||
| Learning_rate: 2e-4 | ||
|---|---|---|
| Data length | ||
| Input ( | Diffusion-step embedding ( | |
| After Conv1x1 ( | After FC: ( | |
| After FC: ( | ||
| Residual block (30 layers) | After FC: ( | |
| Broadcast: ( | ||
| Dilated Conv setting: kernel size: 3, dilation: 2c mod 10 | ||
| After Dilated Conv: ( | ||
| After left Conv1x1: ( | After right Conv1x1: ( | |
| Output | After Conv1x1 ( | |
| After Conv1x1 ( | ||
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.