add mt bench fig
Browse files
README.md
CHANGED
|
@@ -63,9 +63,10 @@ Zamba2-2.7B-Instruct punches dramatically above its weight, achieving extremely
|
|
| 63 |
Moreover, due to its unique hybrid SSM architecture, Zamba2-2.7B-Instruct achieves extremely low inference latency and rapid generation with a significantly smaller memory footprint than comparable transformer-based models.
|
| 64 |
|
| 65 |
<center>
|
| 66 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/
|
| 67 |
</center>
|
| 68 |
|
|
|
|
| 69 |
Time to First Token (TTFT) | Output Generation
|
| 70 |
:-------------------------:|:-------------------------:
|
| 71 |
 | 
|
|
|
|
| 63 |
Moreover, due to its unique hybrid SSM architecture, Zamba2-2.7B-Instruct achieves extremely low inference latency and rapid generation with a significantly smaller memory footprint than comparable transformer-based models.
|
| 64 |
|
| 65 |
<center>
|
| 66 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/65bc13717c6ad1994b6619e9/GyojH0mFCaAaAHBAXlm4T.png" width="700" alt="Zamba performance">
|
| 67 |
</center>
|
| 68 |
|
| 69 |
+
|
| 70 |
Time to First Token (TTFT) | Output Generation
|
| 71 |
:-------------------------:|:-------------------------:
|
| 72 |
 | 
|