Update README.md
Browse files
README.md
CHANGED
|
@@ -62,9 +62,7 @@ print(tokenizer.decode(outputs[0]))
|
|
| 62 |
|
| 63 |
Zamba2-1.2B utilizes and extends our original Zamba hybrid SSM-attention architecture. The core Zamba architecture consists of a backbone of Mamba layers interleaved with one or more shared attention layers (one shared attention in Zamba1, two in Zamba2). This attention has shared weights to minimize the parameter cost of the model. We find that concatenating the original model embeddings to the input to this attention block improves performance, likely due to better maintenance of information across depth. The Zamba2 architecture also applies LoRA projection matrices to the shared transformer blocks to gain some additional expressivity in each block and allow each shared block to specialize slightly to its own unique position while keeping the additional parameter overhead small.
|
| 64 |
|
| 65 |
-
|
| 66 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/65c05e75c084467acab2f84a/XrEIEBxd0fqIgh3LyArAV.png" width="300" alt="Zamba architecture">
|
| 67 |
-
</center>
|
| 68 |
|
| 69 |
|
| 70 |
## Performance
|
|
@@ -73,22 +71,16 @@ Zamba2-1.2B achieves leading and state-of-the-art performance among models of <3
|
|
| 73 |
|
| 74 |
Zamba2-1.2B's high performance and small inference compute and memory footprint renders it an ideal generalist model for on-device applications.
|
| 75 |
|
| 76 |
-
|
| 77 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/65c05e75c084467acab2f84a/U7VD9PYLj3XcEjgV08sP5.png" width="700" alt="Zamba performance">
|
| 78 |
-
</center>
|
| 79 |
|
| 80 |
-
|
| 81 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/65bc13717c6ad1994b6619e9/3u8k7tcRi-oC_ltGhdHAk.png" width="800" alt="Zamba performance">
|
| 82 |
-
</center>
|
| 83 |
|
| 84 |
Time to First Token (TTFT) | Output Generation
|
| 85 |
:-------------------------:|:-------------------------:
|
| 86 |
 | 
|
| 87 |
|
| 88 |
|
| 89 |
-
|
| 90 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/65bc13717c6ad1994b6619e9/nhoss41xlzfEBZzcQXI6z.png" width="700" alt="Zamba inference and memory cost">
|
| 91 |
-
</center>
|
| 92 |
|
| 93 |
## Notice
|
| 94 |
|
|
|
|
| 62 |
|
| 63 |
Zamba2-1.2B utilizes and extends our original Zamba hybrid SSM-attention architecture. The core Zamba architecture consists of a backbone of Mamba layers interleaved with one or more shared attention layers (one shared attention in Zamba1, two in Zamba2). This attention has shared weights to minimize the parameter cost of the model. We find that concatenating the original model embeddings to the input to this attention block improves performance, likely due to better maintenance of information across depth. The Zamba2 architecture also applies LoRA projection matrices to the shared transformer blocks to gain some additional expressivity in each block and allow each shared block to specialize slightly to its own unique position while keeping the additional parameter overhead small.
|
| 64 |
|
| 65 |
+
TODO
|
|
|
|
|
|
|
| 66 |
|
| 67 |
|
| 68 |
## Performance
|
|
|
|
| 71 |
|
| 72 |
Zamba2-1.2B's high performance and small inference compute and memory footprint renders it an ideal generalist model for on-device applications.
|
| 73 |
|
| 74 |
+
TODO
|
|
|
|
|
|
|
| 75 |
|
| 76 |
+
TODO
|
|
|
|
|
|
|
| 77 |
|
| 78 |
Time to First Token (TTFT) | Output Generation
|
| 79 |
:-------------------------:|:-------------------------:
|
| 80 |
 | 
|
| 81 |
|
| 82 |
|
| 83 |
+
TODO
|
|
|
|
|
|
|
| 84 |
|
| 85 |
## Notice
|
| 86 |
|