feat: return from_bert for from_pretrained

modeling_lora.py

@@ -1,4 +1,5 @@
 import math
+import os
 from functools import partial
 from typing import Iterator, Optional, Tuple, Union
 
@@ -6,6 +7,7 @@ import torch
 import torch.nn.utils.parametrize as parametrize
 from torch import nn
 from torch.nn import Parameter
+from transformers import PretrainedConfig
 
 from .modeling_bert import BertModel, BertPreTrainedModel, JinaBertConfig
 
@@ -193,6 +195,25 @@ class BertLoRA(BertPreTrainedModel):
         return cls(config, bert=bert, num_adaptions=num_adaptions)
 
 
+    @classmethod
+    def from_pretrained(
+        cls,
+        pretrained_model_name_or_path: Optional[Union[str, os.PathLike]],
+        *model_args,
+        config: Optional[Union[PretrainedConfig, str, os.PathLike]] = None,
+        cache_dir: Optional[Union[str, os.PathLike]] = None,
+        ignore_mismatched_sizes: bool = False,
+        force_download: bool = False,
+        local_files_only: bool = False,
+        token: Optional[Union[str, bool]] = None,
+        revision: str = "main",
+        use_safetensors: bool = None,
+        **kwargs,
+    ):
+        # TODO: choose between from_bert and super().from_pretrained
+        return cls.from_bert(pretrained_model_name_or_path)
+
+
     def _register_lora(self, num_adaptions=1, rank=4, lora_dropout_p=0.0, lora_alpha=1):
         self.apply(
             partial(