Update README.md
Browse files
README.md
CHANGED
|
@@ -47,6 +47,7 @@
|
|
| 47 |
**Configuration-Specific Shapes**:
|
| 48 |
|
| 49 |
| Input | Default | NVIDIA Low | NVIDIA High |
|
|
|
|
| 50 |
| chunk | `[1, 112, 128]` | `[1, 112, 128]` | `[1, 3048, 128]` |
|
| 51 |
| chunk_lengths | `[1]` | `[1]` | `[1]` |
|
| 52 |
| spkcache | `[1, 188, 512]` | `[1, 188, 512]` | `[1, 188, 512]` |
|
|
@@ -55,12 +56,19 @@
|
|
| 55 |
| fifo_lengths | `[1]` | `[1]` | `[1]` |
|
| 56 |
|
| 57 |
| Output | Default | NVIDIA Low | NVIDIA High |
|
|
|
|
| 58 |
| speaker_preds | `[1, 242, 128]` | `[1, 390, 128]` | `[1, 609, 128]` |
|
| 59 |
| chunk_pre_encoder_embs | `[1, 14, 512]` | `[1, 14, 512]` | `[1, 381, 512]` |
|
| 60 |
| chunk_pre_encoder_lengths | `[1]` | `[1]` | `[1]` |
|
| 61 |
| nest_encoder_embs | `[1, 242, 192]` | `[1, 390, 192]` | `[1, 609, 192]` |
|
| 62 |
| nest_encoder_lengths | `[1]` | `[1]` | `[1]` |
|
| 63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
## Usage with FluidAudio (Swift)
|
| 65 |
|
| 66 |
```swift
|
|
|
|
| 47 |
**Configuration-Specific Shapes**:
|
| 48 |
|
| 49 |
| Input | Default | NVIDIA Low | NVIDIA High |
|
| 50 |
+
|-------|---------|------------|-------------|
|
| 51 |
| chunk | `[1, 112, 128]` | `[1, 112, 128]` | `[1, 3048, 128]` |
|
| 52 |
| chunk_lengths | `[1]` | `[1]` | `[1]` |
|
| 53 |
| spkcache | `[1, 188, 512]` | `[1, 188, 512]` | `[1, 188, 512]` |
|
|
|
|
| 56 |
| fifo_lengths | `[1]` | `[1]` | `[1]` |
|
| 57 |
|
| 58 |
| Output | Default | NVIDIA Low | NVIDIA High |
|
| 59 |
+
|--------|---------|------------|-------------|
|
| 60 |
| speaker_preds | `[1, 242, 128]` | `[1, 390, 128]` | `[1, 609, 128]` |
|
| 61 |
| chunk_pre_encoder_embs | `[1, 14, 512]` | `[1, 14, 512]` | `[1, 381, 512]` |
|
| 62 |
| chunk_pre_encoder_lengths | `[1]` | `[1]` | `[1]` |
|
| 63 |
| nest_encoder_embs | `[1, 242, 192]` | `[1, 390, 192]` | `[1, 609, 192]` |
|
| 64 |
| nest_encoder_lengths | `[1]` | `[1]` | `[1]` |
|
| 65 |
|
| 66 |
+
|
| 67 |
+
| Metric | Default | NVIDIA High |
|
| 68 |
+
|---------------|---------|-------------|
|
| 69 |
+
| Latency | ~1.12s | ~30.4s |
|
| 70 |
+
| RTFx (M4 Max) | ~5.7x | ~125.3x |
|
| 71 |
+
|
| 72 |
## Usage with FluidAudio (Swift)
|
| 73 |
|
| 74 |
```swift
|