Update app.py

app.py

@@ -127,7 +127,7 @@ pipe = load_and_prepare_model()
 
 text_encoder=CLIPTextModel.from_pretrained('ford442/RealVisXL_V5.0_BF16', low_cpu_mem_usage=False, subfolder='text_encoder',token=True)#.to(device=device, dtype=torch.bfloat16)
 text_encoder_2=CLIPTextModelWithProjection.from_pretrained('ford442/RealVisXL_V5.0_BF16', low_cpu_mem_usage=False, subfolder='text_encoder_2',token=True)#.to(device=device, dtype=torch.bfloat16)
-   
+
 MAX_SEED = np.iinfo(np.int32).max
 
 neg_prompt_2 = " 'non-photorealistic':1.5, 'unrealistic skin','unattractive face':1.3, 'low quality':1.1, ('dull color scheme', 'dull colors', 'digital noise':1.2),'amateurish', 'poorly drawn face':1.3, 'poorly drawn', 'distorted face', 'low resolution', 'simplistic' "
@@ -193,22 +193,22 @@ def generate_30(
     pipe.text_encoder=text_encoder.to(device=device, dtype=torch.bfloat16)
     pipe.text_encoder_2=text_encoder_2.to(device=device, dtype=torch.bfloat16)
 
-    text_inputs1 = tokenizer(
+    text_inputs1 = pipe.tokenizer(
                     prompt,
                     padding="max_length",
                     max_length=77,
                     truncation=True,
                     return_tensors="pt",
     )
-    text_inputs2 = tokenizer(
+    text_inputs2 = pipe.tokenizer(
                     prompt2,
                     padding="max_length",
                     max_length=77,
                     truncation=True,
                     return_tensors="pt",
     )
-    prompt_embedsa = text_encoder(text_input_ids.to(device), output_hidden_states=True)
-    prompt_embedsb = text_encoder(text_input_ids.to(device), output_hidden_states=True)
+    prompt_embedsa = pipe.text_encoder(text_input_ids.to(device), output_hidden_states=True)
+    prompt_embedsb = pipe.text_encoder(text_input_ids.to(device), output_hidden_states=True)
     prompt_embeds = torch.cat([prompt_embedsa,prompt_embedsb]).mean(dim=-1)
 
     options = {