写好了ocr_model训练脚本的大致框架

src/models/ocr_model/utils/get.py

@@ -0,0 +1,28 @@
+from ....globals import VOCAB_SIZE
+from typing import (
+    Tuple
+)
+
+from transformers import (
+    RobertaConfig,
+    RobertaModel,
+    RobertaTokenizerFast
+)
+
+
+def get_encoder():
+    ...
+
+
+def get_tokenizer() -> RobertaTokenizerFast:
+    ...
+
+
+def get_decoder() -> RobertaModel:
+    configuration = RobertaConfig(
+        vocab_size=VOCAB_SIZE,
+        is_decoder=True
+    )
+    model = RobertaModel(configuration)
+    return model
+