Update README.md

Fixed position of training data desc.

README.md

@@ -47,9 +47,6 @@ The best checkpoint has been saved and made available for use.
 **NUMBER OF INPUTS:** 2 - The SQL queries as string and extra numeric features.  
 **NUMBER OF OUTPUTS:** 1 - Probability that the given SQL is malicious (the output layer uses a sigmoid activation).
 
-The training data is made available [here](dataset/train.csv) and the benchmark data is made
-available [here](dataset/benchmark.csv). The data was curated from the following sources -
-
 #### Checkpointed Epoch
 
 ```
@@ -67,6 +64,9 @@ available [here](dataset/benchmark.csv). The data was curated from the following
 
 #### Training Data
 
+The training data is made available [here](dataset/train.csv) and the benchmark data is made
+available [here](dataset/benchmark.csv). The data was curated from the following sources -
+
 1. https://www.kaggle.com/datasets/gambleryu/biggest-sql-injection-dataset/data
 2. https://huggingface.co/datasets/b-mc2/sql-create-context
 3. https://github.com/payloadbox/sql-injection-payload-list/tree/master/Intruder