typo

README.md

@@ -13,7 +13,7 @@ More than one training run goes into making a large language model, but develope
 ## 25 Data Recipes
 
 We call the 25 corpora we train on *data recipes* as they range across popular corpora including Dolma, DCLM, RefinedWeb, C4, and FineWeb as well as combinations of interventions
-on these datasets such as source mixing, deduplication, and filtering. This HuggingFace Dataset contains the tokenized data used to build these recipes with as mapped by this [OLMo script](https://github.com/allenai/OLMo/blob/DataDecide/olmo/data/named_data_mixes.py).
+on these datasets such as source mixing, deduplication, and filtering. This HuggingFace Dataset contains the tokenized data used to build these recipes, as mapped by this [OLMo script](https://github.com/allenai/OLMo/blob/DataDecide/olmo/data/named_data_mixes.py).
 
 | **Source**                          | **Recipe**                     | **Description** |
 |------------------------------------|--------------------------------|-----------------|