Update README.md

README.md

@@ -18,7 +18,9 @@ Train Accuracy: 0.9893
 Validation Loss: 0.0155  
 Validation Accuracy: 0.9702  
 Epoch: 20 
-
+## The dataset for training PPPSL-ESM2
+The full dataset contains 11,970 protein sequences, including Cellwall (87), Cytoplasmic (6,905), CYtoplasmic Membrane (2,567), Extracellular (1085), Outer Membrane (758), and Periplasmic (568).
+The highly imbalanced sample sizes across the six categories in this dataset pose a significant challenge for classification.  
 ## Example
 ```
 from transformers import AutoTokenizer, AutoModelForSequenceClassification