Update README.md

README.md

@@ -64,13 +64,8 @@ Zyda also outperforms Dolma, RefinedWeb, and Fineweb on 1.4B models trained on 5
 
 According to our evaluations, Zyda is the most performant per-token open dataset available.
 
-<center>
-    <img src="https://huggingface.co/datasets/Zyphra/Zyda/blob/main/Zyda_dataset_pie.png" alt="Zyda scores across time ablations">
-</center>
 
-<figure style="width: 120%"> <img src="https://huggingface.co/datasets/Zyphra/Zyda/blob/main/Zyda_dataset_pie.png"> <figcaption><em>Zyda scores across time vs other datasets </em></figcaption></figure>
-
-![image/png](https://huggingface.co/datasets/Zyphra/Zyda/blob/main/Zyda_dataset_pie.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/65c05e75c084467acab2f84a/fXaQAOBDJpoaAr1clfTel.png)
 
 ## How to download
 
@@ -84,6 +79,8 @@ For downloading individual components put their name in the name arg of `load_da
 - zyda_arxiv_only
 - zyda_c4-en_only
 - zyda_peS2o_only
+
+
 - zyda_pile-uncopyrighted_only
 - zyda_refinedweb_only
 - zyda_slimpajama_only
@@ -110,6 +107,7 @@ Dataset fields:
 - `source_other`: metadata from the source dataset (converted to json string)
 
 
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/65c05e75c084467acab2f84a/dQV8zNTNCx1xMMT-iupY6.png)
 
 ### Source Data