Create README.md

README.md

@@ -0,0 +1,24 @@
+This model was trained with MLSUM (Turkish language) summarization datasets where I fine-tuned google/mt5-small using [SimpleT5](https://github.com/Shivanandroy/simpleT5) library.
+The first results are not promising may be due to using small check-points. I will work on it for improvements!
+
+The code piece for training
+
+```
+from simplet5 import SimpleT5
+model = SimpleT5()
+model.from_pretrained("mt5","google/mt5-small")
+
+# train
+model.train(train_df=train2, # pandas dataframe with 2 columns: source_text & target_text
+            eval_df=validation2, # pandas dataframe with 2 columns: source_text & target_text
+            source_max_token_len = 512, 
+            target_max_token_len = 128,
+            batch_size = 8,
+            max_epochs = 5,
+            use_gpu = True,
+            outputdir = "mt5_mlsum_turkish",
+            early_stopping_patience_epochs = 0,
+            precision = 32
+)
+```
+