Update README.md
Browse files
README.md
CHANGED
|
@@ -1,6 +1,8 @@
|
|
| 1 |
---
|
| 2 |
library_name: transformers
|
| 3 |
-
|
|
|
|
|
|
|
| 4 |
---
|
| 5 |
|
| 6 |
# Model Card for Model ID
|
|
@@ -9,8 +11,10 @@ tags: []
|
|
| 9 |
|
| 10 |
|
| 11 |
|
| 12 |
-
## Model
|
| 13 |
-
|
|
|
|
|
|
|
| 14 |
### Model Description
|
| 15 |
|
| 16 |
<!-- Provide a longer summary of what this model is. -->
|
|
|
|
| 1 |
---
|
| 2 |
library_name: transformers
|
| 3 |
+
license: mit
|
| 4 |
+
datasets:
|
| 5 |
+
- lmms-lab/DocVQA
|
| 6 |
---
|
| 7 |
|
| 8 |
# Model Card for Model ID
|
|
|
|
| 11 |
|
| 12 |
|
| 13 |
|
| 14 |
+
## Model Description
|
| 15 |
+
DIVE-Doc is an end-to-end trade-off between LVLMs and lightweight architectures in the context of DocVQA.
|
| 16 |
+
It is built by distilling the [SigLIP-400m](https://arxiv.org/abs/2303.15343) visual encoder of [PaliGEMMA]() into a small hierarchical [Swin transformer](), while reusing the original [GEMMA]() decoder.
|
| 17 |
+
This allowed DIVE-Doc to keep competitive performance with its teacher while reducing the visual encoder's parameters to one-fifth.
|
| 18 |
### Model Description
|
| 19 |
|
| 20 |
<!-- Provide a longer summary of what this model is. -->
|