Update app.py

app.py

@@ -74,7 +74,7 @@ def generate_image(prompt,
                                                                sft_format=processor.sft_format,
                                                                system_prompt='')
     text = text + processor.image_start_tag
-    input_ids = torch.LongTensor(processor.tokenizer.encode(prompt))
+    input_ids = torch.LongTensor(processor.tokenizer.encode(text))
     output, patches = generate(input_ids,
                                width // 16 * 16,
                                height // 16 * 16,
@@ -89,8 +89,8 @@ with gr.Blocks() as demo:
     with gr.Row():
         with gr.Column():
             prompt = gr.Textbox(label='Prompt', value='portrait, color, cinematic')
-            width = gr.Slider(128, 1536, 128, step=16, label='Width')
-            height = gr.Slider(128, 1536, 128, step=16, label='Height')
+            width = gr.Slider(64, 1536, 384, step=16, label='Width')
+            height = gr.Slider(64, 1536, 384, step=16, label='Height')
             guidance = gr.Slider(1.0, 10.0, 5, step=0.1, label='Guidance')
             seed = gr.Number(-1, precision=0, label='Seed (-1 for random)')
 
@@ -119,8 +119,11 @@ if __name__ == '__main__':
     tokenizer = processor.tokenizer
     # model: MultiModalityCausalLM = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
     config = AutoConfig.from_pretrained(model_path)
-    config.language_config._attn_implementation = 'eager'
-    model = AutoModelForCausalLM.from_config(config, trust_remote_code=True)
+    language_config = config.language_config
+    language_config._attn_implementation = 'eager'
+    model = AutoModelForCausalLM.from_pretrained(model_path,
+                                                 language_config=language_config,
+                                                 trust_remote_code=True)
     if torch.cuda.is_available():
         model = model.to(torch.bfloat16).cuda()
     else: