Update README.md
Browse files
README.md
CHANGED
|
@@ -37,11 +37,6 @@ Data is the same as the ColPali data described in the paper.
|
|
| 37 |
|
| 38 |
## Model Training
|
| 39 |
|
| 40 |
-
### Dataset
|
| 41 |
-
The dataset was extended from the original colpali train set with the gemini 1.5 flash generated QA on 35k images scraped from internet.
|
| 42 |
-
|
| 43 |
-
*Note: Multilingual data is present in the pretraining corpus of the language model and most probably in the multimodal training.*
|
| 44 |
-
|
| 45 |
### Parameters
|
| 46 |
We train models use low-rank adapters ([LoRA](https://arxiv.org/abs/2106.09685))
|
| 47 |
with `alpha=128` and `r=128` on the transformer layers from the language model,
|
|
|
|
| 37 |
|
| 38 |
## Model Training
|
| 39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
### Parameters
|
| 41 |
We train models use low-rank adapters ([LoRA](https://arxiv.org/abs/2106.09685))
|
| 42 |
with `alpha=128` and `r=128` on the transformer layers from the language model,
|