Update README.md

README.md

@@ -0,0 +1,10 @@
+## TextAttack Model Card
+This `albert-base-v2` model was fine-tuned for sequence classification using TextAttack 
+and the glue dataset loaded using the `nlp` library. The model was fine-tuned 
+for 5 epochs with a batch size of 32, a learning 
+rate of 3e-05, and a maximum sequence length of 128. 
+Since this was a regression task, the model was trained with a mean squared error loss function. 
+The best score the model achieved on this task was 0.9064220351504577, as measured by the 
+eval set pearson correlation, found after 3 epochs.
+
+For more information, check out [TextAttack on Github](https://github.com/QData/TextAttack).