Video-Text-to-Text
Transformers
Safetensors
English
gemma4
image-text-to-text
video-captioning
multimodal
gemma
parakeet
Instructions to use SulphurAI/sulphur-caption with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SulphurAI/sulphur-caption with Transformers:
# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("SulphurAI/sulphur-caption") model = AutoModelForMultimodalLM.from_pretrained("SulphurAI/sulphur-caption", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -91,6 +91,22 @@ The model was trained with some settings, grouped under `Prompt Settings`. Here
|
|
| 91 |
- `CAMERA_DETAIL` | "medium"
|
| 92 |
- `CAPTION_STYLE` | "plain"
|
| 93 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
Runtime and generation settings are grouped under `vLLM / Generation Hyperparameters`, including:
|
| 95 |
|
| 96 |
- `NUM_FRAMES`
|
|
|
|
| 91 |
- `CAMERA_DETAIL` | "medium"
|
| 92 |
- `CAPTION_STYLE` | "plain"
|
| 93 |
|
| 94 |
+
### Possible Options
|
| 95 |
+
|
| 96 |
+
- `HAS_THINKING` | True, False
|
| 97 |
+
- `CAPTION_LENGTH` | "very small", "small", "medium", "large", "very large"
|
| 98 |
+
- `INCLUDE_WATERMARK_INFO` | True, False
|
| 99 |
+
- `VULGARITY` | "none", "low", "medium", "high"
|
| 100 |
+
- `UNCERTAINTY` | "none", "low", "medium", "high"
|
| 101 |
+
- `CHARACTER_NAMES` | "none", "ambiguous", "single", "multiple"
|
| 102 |
+
- `FLUFF` | "none", "low", "medium", "high"
|
| 103 |
+
- `HAS_REPETITION` | True, False
|
| 104 |
+
- `SPECULATION` | "none", "low", "medium", "high"
|
| 105 |
+
- `TEMPORAL_DETAIL` | "static", "low", "medium", "high"
|
| 106 |
+
- `VISUAL_SPECIFICITY` | "generic", "moderate", "detailed", "excessive"
|
| 107 |
+
- `CAMERA_DETAIL` | "none", "low", "medium", "high"
|
| 108 |
+
- `CAPTION_STYLE` | "plain", "verbose", "ornate", "robotic"
|
| 109 |
+
|
| 110 |
Runtime and generation settings are grouped under `vLLM / Generation Hyperparameters`, including:
|
| 111 |
|
| 112 |
- `NUM_FRAMES`
|