Upload anytext.py

anytext.py

@@ -533,9 +533,9 @@ class AbstractEncoder(nn.Module):
         raise NotImplementedError
 
 
-class FrozenCLIPEmbedderT3(AbstractEncoder, nn.Module):
+class FrozenCLIPEmbedderT3(AbstractEncoder, ModelMixin, ConfigMixin):
     """Uses the CLIP transformer encoder for text (from Hugging Face)"""
-
+    @register_to_config
     def __init__(
         self,
         device="cpu",
@@ -547,8 +547,8 @@ class FrozenCLIPEmbedderT3(AbstractEncoder, nn.Module):
         super().__init__()
         self.tokenizer = CLIPTokenizer.from_pretrained("tolgacangoz/anytext", subfolder="tokenizer")
         self.transformer = CLIPTextModel.from_pretrained("tolgacangoz/anytext", subfolder="text_encoder")
-        self.device = device
-        self.max_length = max_length
+        # self.device = device
+        # self.max_length = max_length
         if freeze:
             self.freeze()
 
@@ -727,10 +727,10 @@ class FrozenCLIPEmbedderT3(AbstractEncoder, nn.Module):
             tokens_list.append(remaining_group_pad)
         return tokens_list
 
-    def to(self, *args, **kwargs):
-        self.transformer = self.transformer.to(*args, **kwargs)
-        self.device = self.transformer.device
-        return self
+    # def to(self, *args, **kwargs):
+    #     self.transformer = self.transformer.to(*args, **kwargs)
+    #     self.device = self.transformer.device
+    #     return self
 
 
 class TextEmbeddingModule(ModelMixin, ConfigMixin):