updates

README.md

@@ -32,7 +32,7 @@ Read more about some of the technical challenges encountered during training IDE
 - **Developed by:** Hugging Face
 - **Model type:** Multi-modal model (image+text)
 - **Language(s) (NLP):** en
-- **License:** other
+- **License:** see [License section](#license)
 - **Parent Model:** [laion/CLIP-ViT-H-14-laion2B-s32B-b79K](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K) and [huggyllama/llama-65b](https://huggingface.co/huggyllama/llama-65b)
 - **Resources for more information:**
     - [GitHub Repo](https://github.com/huggingface/m4/)
@@ -158,7 +158,9 @@ We perform checkpoint selection based on validation sets of VQAv2, TextVQA, OKVQ
 
 As opposed to Flamingo, we did not train IDEFICS on video-text pairs datasets, and as such, we did not evaluate the model on video-text benchmarks like Flamingo did. We leave that evaluation for a future iteration.
 
-<!-- <img src="./assets/Figure_Evals_IDEFIX.png"  width="55%"> <img width=120/> -->
+<img src="./assets/Figure_Evals_IDEFIX.png"  width="55%">
+
+We note that since IDEFICS was trained on PMD (which contains COCO), the evaluation numbers on COCO are not directly comparable with Flamingo and OpenFlamingo since they did not explicitely have this dataset in the training mixture. Additionally, Flamingo is trained with images of resolution 320 x 320 while IDEFICS and OpenFlamingo were trained with images of 224 x 224 resolution.
 
 | Model      |   Shots | VQAv2<br>OE VQA acc.<br> | OKVQA<br>OE VQA acc.<br> | TextVQA<br>OE VQA acc.<br> | VizWiz<br>OE VQA acc.<br> |  TextCaps<br>CIDEr<br> | Coco<br>CIDEr<br> | NoCaps<br>CIDEr | Flickr<br>CIDEr | VisDial<br>NDCG | HatefulMemes<br>ROC AUC | ScienceQA<br>acc. |  RenderedSST2<br>acc. |   Winoground<br>group (text/image) |
 |:-----------|--------:|---------------------:|---------------------:|-----------------------:|----------------------:|-------------------:|---------------:|-----------------:|-----------------:|-----------------:|-------------------------:|-----------------------:|--------------------------:|----------------------------------:|
@@ -174,8 +176,6 @@ As opposed to Flamingo, we did not train IDEFICS on video-text pairs datasets, a
 |            |      16 |                 57.0 |                 48.4 |                   27.9 |                  42.6 |               67.4 |           99.7 |             89.4 |             64.5 |             - |                     50.9 |                   - |                      67.8 |                              - |
 |            |      32 |                 57.9 |                 49.6 |                   28.3 |                  43.7 |               68.1 |           98.0 |             90.5 |             64.4 |             - |                     49.8 |                   - |                      67.0 |                              - |
 
-We note that since we trained on PMD which contains COCO, the evaluation numbers on COCO are not directly comparable with Flamingo and OpenFlamingo since they did not explicitely have this dataset in the training mixture.
-
 For ImageNet-1k, we also report results where the priming samples are selected to be similar (i.e. close in a vector space) to the queried instance.
 
 ImageNet-1k Evaluation: