Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -64,12 +64,12 @@ called the scene "distorted and warped", which was a fair description of what it
|
|
| 64 |
given.
|
| 65 |
|
| 66 |
```bash
|
| 67 |
-
python convert/
|
| 68 |
```
|
| 69 |
|
| 70 |
## Conversion
|
| 71 |
|
| 72 |
-
`convert/
|
| 73 |
|
| 74 |
- **SigLIP 2 reads its grid out of a tensor** to size the position embeddings, which
|
| 75 |
`torch.export` cannot follow. The grid is fixed here, so the resize is computed once and
|
|
|
|
| 64 |
given.
|
| 65 |
|
| 66 |
```bash
|
| 67 |
+
python convert/run_vl.py <image> "What is in this picture?"
|
| 68 |
```
|
| 69 |
|
| 70 |
## Conversion
|
| 71 |
|
| 72 |
+
`convert/export_vl_bundle.py`. Four things needed re-authoring:
|
| 73 |
|
| 74 |
- **SigLIP 2 reads its grid out of a tensor** to size the position embeddings, which
|
| 75 |
`torch.export` cannot follow. The grid is fixed here, so the resize is computed once and
|