Update README.md
Browse files
README.md
CHANGED
|
@@ -125,12 +125,6 @@ Supervised speech instruction finetuning via knowledge-distillation. For more in
|
|
| 125 |
- **Training regime:** BF16 mixed precision training
|
| 126 |
- **Hardward used:** 8x H100 GPUs
|
| 127 |
|
| 128 |
-
#### Speeds, Sizes, Times
|
| 129 |
-
|
| 130 |
-
The current version of Ultravox, when invoked with audio content, has a time-to-first-token (TTFT) of approximately 150ms, and a tokens-per-second rate of ~50-100 when using an A100-40GB GPU, all using a Llama 3.3 70B backbone.
|
| 131 |
-
|
| 132 |
-
Check out the audio tab on [TheFastest.ai](https://thefastest.ai/?m=audio) for daily benchmarks and a comparison with other existing models.
|
| 133 |
-
|
| 134 |
## Evaluation
|
| 135 |
|
| 136 |
| | Ultravox 0.4 70B | Ultravox 0.4.1 70B | **Ultravox 0.5 70B** |
|
|
|
|
| 125 |
- **Training regime:** BF16 mixed precision training
|
| 126 |
- **Hardward used:** 8x H100 GPUs
|
| 127 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
## Evaluation
|
| 129 |
|
| 130 |
| | Ultravox 0.4 70B | Ultravox 0.4.1 70B | **Ultravox 0.5 70B** |
|