Update weights and modeling file

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58757d657ff44051314c8030b68e04cb1bb618ca9a4885418f111f6fb708185a
+size 463221266

modeling_florence2.py

@@ -26,9 +26,10 @@ import torch.utils.checkpoint as checkpoint
 from torch.nn import CrossEntropyLoss 
 from collections import OrderedDict
 from einops import rearrange
-from timm.models.layers import DropPath, trunc_normal_
+from timm.layers import DropPath, trunc_normal_
 
 from transformers.modeling_utils import PreTrainedModel
+from transformers.generation.utils import GenerationMixin
 from transformers.utils import (
     ModelOutput,
     add_start_docstrings,
@@ -609,29 +610,10 @@ class DaViT(nn.Module):
         self.avgpool = nn.AdaptiveAvgPool1d(1)
         self.head = nn.Linear(self.embed_dims[-1], num_classes) if num_classes > 0 else nn.Identity()
 
-        self.apply(self._init_weights)
-
     @property
     def dim_out(self):
         return self.embed_dims[-1]
 
-    def _init_weights(self, m):
-        if isinstance(m, nn.Linear):
-            trunc_normal_(m.weight, std=0.02)
-            if m.bias is not None:
-                nn.init.constant_(m.bias, 0)
-        elif isinstance(m, nn.Conv2d):
-            nn.init.normal_(m.weight, std=0.02)
-            for name, _ in m.named_parameters():
-                if name in ['bias']:
-                    nn.init.constant_(m.bias, 0)
-        elif isinstance(m, nn.LayerNorm):
-            nn.init.constant_(m.weight, 1.0)
-            nn.init.constant_(m.bias, 0)
-        elif isinstance(m, nn.BatchNorm2d):
-            nn.init.constant_(m.weight, 1.0)
-            nn.init.constant_(m.bias, 0)
-
     def forward_features_unpool(self, x):
         """
         forward until avg pooling 
@@ -1450,6 +1432,17 @@ class Florence2LanguagePreTrainedModel(PreTrainedModel):
             module.weight.data.normal_(mean=0.0, std=std)
             if module.padding_idx is not None:
                 module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.Conv2d):
+            nn.init.normal_(module.weight, std=0.02)
+            for name, _ in module.named_parameters():
+                if name == "bias":
+                    nn.init.constant_(module.bias, 0)
+        elif isinstance(module, nn.LayerNorm):
+            nn.init.constant_(module.weight, 1.0)
+            nn.init.constant_(module.bias, 0)
+        elif isinstance(module, nn.BatchNorm2d):
+            nn.init.constant_(module.weight, 1.0)
+            nn.init.constant_(module.bias, 0)
 
     @property
     def dummy_inputs(self):
@@ -2059,7 +2052,7 @@ class Florence2LanguageModel(Florence2LanguagePreTrainedModel):
         )
 
 
-class Florence2LanguageForConditionalGeneration(Florence2LanguagePreTrainedModel):
+class Florence2LanguageForConditionalGeneration(Florence2LanguagePreTrainedModel, GenerationMixin):
     base_model_prefix = "model"
     _tied_weights_keys = ["encoder.embed_tokens.weight", "decoder.embed_tokens.weight", "lm_head.weight"]
     _keys_to_ignore_on_load_missing = ["final_logits_bias"]