Add PyTorch model

flax_to_torch.py

@@ -0,0 +1,22 @@
+import argparse
+
+from transformers import RobertaForMaskedLM
+
+
+def convert_flax_model_to_torch(flax_model_path: str, torch_model_path: str = "./"):
+    """
+    Converts Flax model weights to PyTorch weights.
+    """
+    model = RobertaForMaskedLM.from_pretrained(flax_model_path, from_flax=True)
+    model.save_pretrained(torch_model_path)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Flax to Pytorch model coversion")
+    parser.add_argument(
+        "--flax_model_path", type=str, default="flax-community/roberta-base-mr", help="Flax model path"
+    )
+    parser.add_argument("--torch_model_path", type=str, default="./", help="PyTorch model path")
+    args = parser.parse_args()
+    convert_flax_model_to_torch(args.flax_model_path, args.torch_model_path)

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b5417d369e470bb94982e578c90cc7d0ad1ec943577b8f92f9918fbe3a2ca2f
+size 498872747