Update README.md
Browse files
README.md
CHANGED
|
@@ -43,7 +43,7 @@ The training data is 100% synthetic and was created by the following pipeline:
|
|
| 43 |
|
| 44 |
1. Randomly roll keywords
|
| 45 |
2. Generate image based on keywords
|
| 46 |
-
3. Caption Image using Qwen3-VL-
|
| 47 |
4. Refine Caption with gpt-oss-120b
|
| 48 |
5. Create training pairs by reversing caption into keywords and one sentence descriptions
|
| 49 |
|
|
|
|
| 43 |
|
| 44 |
1. Randomly roll keywords
|
| 45 |
2. Generate image based on keywords
|
| 46 |
+
3. Caption Image using Qwen3-VL-32b-Instruct-FP8
|
| 47 |
4. Refine Caption with gpt-oss-120b
|
| 48 |
5. Create training pairs by reversing caption into keywords and one sentence descriptions
|
| 49 |
|