Update ldm/modules/encoders/modules.py

ldm/modules/encoders/modules.py

@@ -4,7 +4,7 @@ from torch.utils.checkpoint import checkpoint
 
 from transformers import T5Tokenizer, T5EncoderModel, CLIPTokenizer, CLIPTextModel
 
-from open_clip import get_tokenizer
+import open_clip
 from ldm.util import default, count_params
 from huggingface_hub import hf_hub_download
 
@@ -151,7 +151,7 @@ class FrozenOpenCLIPEmbedder(AbstractEncoder):
         )
 
         self.model = open_clip.create_model(arch, pretrained=pretrained_path)
-        self.tokenizer = get_tokenizer(arch)
+        self.tokenizer = open_clip.tokenizer.SimpleTokenizer()
 
         if hasattr(self.model, "visual"):
             delattr(self.model, "visual")
@@ -174,8 +174,9 @@ class FrozenOpenCLIPEmbedder(AbstractEncoder):
             param.requires_grad = False
 
     def forward(self, text):
-        tokens = open_clip.tokenize(text)
-        z = self.encode_with_transformer(tokens.to(self.device))
+        tokens = [self.tokenizer.encode(t) for t in text]
+        tokens = torch.tensor(tokens).to(self.device)
+        z = self.encode_with_transformer(tokens)
         return z
 
     def encode_with_transformer(self, text):