Update app.py

app.py

@@ -7,15 +7,15 @@ import subprocess
 subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
 models = {
-    "microsoft/Phi-3.5-vision-instruct": AutoModelForCausalLM.from_pretrained("microsoft/Phi-3.5-vision-instruct", trust_remote_code=True, torch_dtype="auto", _attn_implementation="flash_attention_2").cuda().eval()
+    "Qwen/Qwen2-VL-2B-Instruct": AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True, torch_dtype="auto", _attn_implementation="flash_attention_2").cuda().eval()
 
 }
 
 processors = {
-    "microsoft/Phi-3.5-vision-instruct": AutoProcessor.from_pretrained("microsoft/Phi-3.5-vision-instruct", trust_remote_code=True)
+    "Qwen/Qwen2-VL-2B-Instruct": AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True)
 }
 
-DESCRIPTION = "[Phi-3.5-vision Demo](https://huggingface.co/microsoft/Phi-3.5-vision-instruct)"
+DESCRIPTION = "[Qwen2-VL-2B Demo](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct)"
 
 kwargs = {}
 kwargs['torch_dtype'] = torch.bfloat16
@@ -25,23 +25,49 @@ assistant_prompt = '<|assistant|>\n'
 prompt_suffix = "<|end|>\n"
 
 @spaces.GPU
-def run_example(image, text_input=None, model_id="microsoft/Phi-3.5-vision-instruct"):
+def run_example(image, text_input=None, model_id="Qwen/Qwen2-VL-2B-Instruct"):
     model = models[model_id]
     processor = processors[model_id]
 
     prompt = f"{user_prompt}<|image_1|>\n{text_input}{prompt_suffix}{assistant_prompt}"
     image = Image.fromarray(image).convert("RGB")
-
-    inputs = processor(prompt, image, return_tensors="pt").to("cuda:0")
-    generate_ids = model.generate(**inputs, 
-                                max_new_tokens=1000,
-                                eos_token_id=processor.tokenizer.eos_token_id,
-                                )
-    generate_ids = generate_ids[:, inputs['input_ids'].shape[1]:]
-    response = processor.batch_decode(generate_ids, 
-                                    skip_special_tokens=True, 
-                                    clean_up_tokenization_spaces=False)[0]
-    return response
+    messages = [
+    {
+            "role": "user",
+            "content": [
+                {
+                    "type": "image",
+                    "image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
+                },
+                {"type": "text", "text": "Describe this image."},
+            ],
+        }
+    ]
+    
+    # Preparation for inference
+    text = processor.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+    image_inputs, video_inputs = process_vision_info(messages)
+    inputs = processor(
+        text=[text],
+        images=image_inputs,
+        videos=video_inputs,
+        padding=True,
+        return_tensors="pt",
+    )
+    inputs = inputs.to("cuda")
+    
+    # Inference: Generation of the output
+    generated_ids = model.generate(**inputs, max_new_tokens=128)
+    generated_ids_trimmed = [
+        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
+    ]
+    output_text = processor.batch_decode(
+        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
+    )
+    
+    return output_text
 
 css = """
   #output {
@@ -53,11 +79,11 @@ css = """
 
 with gr.Blocks(css=css) as demo:
     gr.Markdown(DESCRIPTION)
-    with gr.Tab(label="Phi-3.5 Input"):
+    with gr.Tab(label="Qwen2-VL-2B Input"):
         with gr.Row():
             with gr.Column():
                 input_img = gr.Image(label="Input Picture")
-                model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="microsoft/Phi-3.5-vision-instruct")
+                model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="Qwen/Qwen2-VL-2B-Instruct")
                 text_input = gr.Textbox(label="Question")
                 submit_btn = gr.Button(value="Submit")
             with gr.Column():