Document PyTorch checkpoint
Browse files
README.md
CHANGED
|
@@ -26,10 +26,12 @@ It does **not** contain the language-model weights or a complete Kimi-VL model.
|
|
| 26 |
|
| 27 |
## Checkpoint
|
| 28 |
|
| 29 |
-
- File: `model.
|
| 30 |
- Source: `moonshotai/Kimi-VL-A3B-Instruct`
|
| 31 |
- Contents: vision tower and multimodal projector only
|
| 32 |
-
- Format:
|
|
|
|
|
|
|
| 33 |
|
| 34 |
## License and attribution
|
| 35 |
|
|
|
|
| 26 |
|
| 27 |
## Checkpoint
|
| 28 |
|
| 29 |
+
- File: `model.pt`
|
| 30 |
- Source: `moonshotai/Kimi-VL-A3B-Instruct`
|
| 31 |
- Contents: vision tower and multimodal projector only
|
| 32 |
+
- Format: PyTorch state dictionary
|
| 33 |
+
- Conversion: resizes the original 64 x 64 positional-embedding grid to the
|
| 34 |
+
current Kornia model's configured grid
|
| 35 |
|
| 36 |
## License and attribution
|
| 37 |
|