update the pre-training checkpoint.

The earlier checkpoint for vilbert 6 layer is pretrained with fixed Bert weight. The new one can replicate the paper performance.

README.md

@@ -40,10 +40,10 @@ Check `README.md` under `data` for more details.  Check  `vlbert_tasks.yml` for
 
 | Model | Objective | Link |
 |:-------:|:------:|:------:|
-|ViLBERT 2-Layer| Conceptual Caption |[Google Drive]()|
-|ViLBERT 4-Layer| Conceptual Caption |[Google Drive]()|
-|ViLBERT 6-Layer| Conceptual Caption |[Google Drive](https://drive.google.com/drive/folders/1Re0L75uazH3Qrep_aRgtaVelDEz4HV9c?usp=sharing)|
-|ViLBERT 8-Layer| Conceptual Caption |[Google Drive]()|
+|ViLBERT 2-Layer| Conceptual Caption |[Google Drive](https://drive.google.com/drive/folders/1asaQDYTacetm12j1K4AkYWdjPincLnap?usp=sharing)|
+|ViLBERT 4-Layer| Conceptual Caption |[Google Drive](https://drive.google.com/drive/folders/1uDa1UsJC-Vz0ZdbHUMw5imRGhk5oM-YR?usp=sharing)|
+|ViLBERT 6-Layer| Conceptual Caption |[Google Drive](https://drive.google.com/drive/folders/1JVM5WiolJJLnY9_lruxSaSop7IFX8a-v?usp=sharing)|
+|ViLBERT 8-Layer| Conceptual Caption |[Google Drive](https://drive.google.com/drive/folders/1M-QoxLB6WJaqY9nq4KzPwfpJ8Va5FNCy?usp=sharing)|
 |ViLBERT 6-Layer| VQA |[Google Drive](https://drive.google.com/drive/folders/1nrcVww0u_vozcFRQVr58-YH5LOU1ZiWT?usp=sharing)|
 |ViLBERT 6-Layer| VCR |[Google Drive](https://drive.google.com/drive/folders/1QJuMzBarTKU_hAWDSZm60rWiDnbAVEVZ?usp=sharing)|
 |ViLBERT 6-Layer| RefCOCO+ |[Google Drive](https://drive.google.com/drive/folders/1GWY2fEbZCYHkcnxd0oysU0olfPdzcD3l?usp=sharing)|