diff --git a/.gitignore b/.gitignore
index 9531f03..ac51a34 100644
--- a/.gitignore
+++ b/.gitignore
@@ -128,3 +128,5 @@ dmypy.json
 
 # modelset test
 modelset
+/ecore.jsonl
+/uml.jsonl
diff --git a/requirements.txt b/requirements.txt
index 3bd6124..cf06bc0 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -5,3 +5,4 @@ scikit-learn
 numpy
 tqdm
 gensim==4.2.0
+datasets
diff --git a/setup.py b/setup.py
index fd9c07c..c07e3a5 100644
--- a/setup.py
+++ b/setup.py
@@ -30,6 +30,7 @@
         "scikit-learn",
         "numpy",
         "tqdm",
-        "gensim==4.2.0"
+        "gensim==4.2.0",
+        "datasets"
     ]
 )
diff --git a/upload_hg/generate_json_files.py b/upload_hg/generate_json_files.py
new file mode 100644
index 0000000..e1fb9b6
--- /dev/null
+++ b/upload_hg/generate_json_files.py
@@ -0,0 +1,63 @@
+import sys
+
+import pandas as pd
+
+sys.path.append("./src")
+
+
+def main():
+    from modelset import load
+    for model_type in ['ecore', 'uml']:
+        dataset = load(modeltype=model_type, selected_analysis=['stats'])
+        dataset_df = dataset.to_normalized_df(min_occurrences_per_category=10)
+        dataset_df_no_dups = dataset.to_normalized_df(remove_duplicates=True,
+                                                      min_occurrences_per_category=10)
+
+        ids = list(dataset_df['id'])
+        ids_no_dups = list(dataset_df_no_dups['id'])
+        print(f'Full dataset {len(ids)}')
+        print(f'No dup dataset {len(ids_no_dups)}')
+        labels = list(dataset_df['category'])
+
+        txt_filenames = [dataset.txt_file(i) for i in ids]
+        txt_contents = []
+        for f in txt_filenames:
+            with open(f, 'r') as file:
+                data = file.read()
+                txt_contents.append(data)
+
+        graph_filenames = [dataset.graph_file(i) for i in ids]
+        graph_contents = []
+        for f in graph_filenames:
+            with open(f, 'r') as file:
+                data = file.read()
+                graph_contents.append(data)
+
+        # Important: Actually, if True, the model does not belong in the deduplicated version
+        is_dup = [True if i not in ids_no_dups else False for i in ids]
+        print(f'Duplicate: {len([f for f in is_dup if f==True])}')
+        print(f'No Duplicate: {len([f for f in is_dup if f == False])}')
+
+        # XMI
+        xmi_files = [dataset.model_file(dataset.get_model_by_id(i)) for i in ids]
+        xmi_contents = []
+        for f in xmi_files:
+            with open(f, 'r') as file:
+                data = file.read()
+                xmi_contents.append(data)
+
+        final_pd = pd.DataFrame.from_dict({
+            "ids": ids,
+            "labels": labels,
+            "txt": txt_contents,
+            "graph": graph_contents,
+            "xmi": xmi_contents,
+            "model_type": [model_type for _ in range(len(ids))],
+            "is_duplicated": is_dup
+        })
+
+        final_pd.to_json(f'{model_type}.jsonl', orient='records')
+
+
+if __name__ == '__main__':
+    main()
diff --git a/upload_hg/test_hg.py b/upload_hg/test_hg.py
new file mode 100644
index 0000000..816a067
--- /dev/null
+++ b/upload_hg/test_hg.py
@@ -0,0 +1,7 @@
+from datasets import load_dataset
+import json
+
+dataset_hg = load_dataset('antolin/modelset', split="train")
+
+print(dataset_hg)
+print(dataset_hg["xmi"][0])