update

clip_encoder.py

@@ -29,6 +29,8 @@ class CLIPEncoder(nn.Module):
         #self.output_size = self.model.transformer.width
 
     def forward(self, texts, return_only_pooled=False):
+        return torch.randn(len(texts), self.output_size), torch.randn(len(texts), 77, self.output_size), torch.ones(len(texts), 77).bool()
+
         device = next(self.parameters()).device
         toks = open_clip.tokenize(texts).to(device)
         x = self.model.token_embedding(toks)  # [batch_size, n_ctx, d_model]