Update README.md
Browse files
README.md
CHANGED
|
@@ -21,8 +21,9 @@ ViLaH (Vision Language Hindi) is a model with 3 billion parameters, fine-tuned f
|
|
| 21 |
* Model Configuration: Fine-tuned on a single epoch using a V100 gpu.
|
| 22 |
* Training Duration: Approximately one day.
|
| 23 |
* Evaluation Loss: Achieved an eval loss of 1.6384 at the end of the epoch.
|
| 24 |
-
|
| 25 |
-
|
|
|
|
| 26 |
# Dataset
|
| 27 |
The dataset was finetuned on only one dataset
|
| 28 |
* [damerajee/clean_hin_vqa](https://huggingface.co/datasets/damerajee/clean_hin_vqa) : This dataset was derived from [Lin-Chen/ShareGPT4V](https://huggingface.co/google/paligemma-3b-pt-224) and filtered to include only images from the COCO dataset. The original dataset was translated and cleaned to ensure high-quality Hindi visual question answering content.
|
|
|
|
| 21 |
* Model Configuration: Fine-tuned on a single epoch using a V100 gpu.
|
| 22 |
* Training Duration: Approximately one day.
|
| 23 |
* Evaluation Loss: Achieved an eval loss of 1.6384 at the end of the epoch.
|
| 24 |
+
* The model is still being train as of right now with better quality dataset
|
| 25 |
+
* The model's performance may be compromised due to insufficient data and the fact that it was trained for only one epoch.
|
| 26 |
+
|
| 27 |
# Dataset
|
| 28 |
The dataset was finetuned on only one dataset
|
| 29 |
* [damerajee/clean_hin_vqa](https://huggingface.co/datasets/damerajee/clean_hin_vqa) : This dataset was derived from [Lin-Chen/ShareGPT4V](https://huggingface.co/google/paligemma-3b-pt-224) and filtered to include only images from the COCO dataset. The original dataset was translated and cleaned to ensure high-quality Hindi visual question answering content.
|