Update README.md

README.md

@@ -16,8 +16,7 @@ The model was pre-trained on a English and Dutch mC4 cleaned.
 
 ## Finetuning
 
-The model was finetuned on **128-max token length** ccmatrix.
-Validated on tatoeba.
-
-Note: multi-direction. Prepend either `translate Dutch to English: `
-or `translate English to Dutch: `
+The model was finetuned on CCMatrix, validated on Tatoeba.
+ * **128-max token length** 
+ * Only the first 25M sentences of CCMatrix were used, both en->nl and en->nl (total 50M sentences).
+ * Note: multi-direction. Prepend either `translate Dutch to English: ` or `translate English to Dutch: `