Change config

app.py

@@ -1,12 +1,16 @@
 from vllm import LLM, SamplingParams
 import gradio as gr
 repo_id = "mistral-community/pixtral-12b-240910" #Replace to the model you would like to use
-sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
+sampling_params = SamplingParams(max_tokens=8192, temperature=0.7)
+max_tokens_per_img = 4096
+max_img_per_msg = 5
 
 # @spaces.GPU #[uncomment to use ZeroGPU]
 def infer(image_url, prompt, progress=gr.Progress(track_tqdm=True)):
     # tokenize image urls and text
-    llm = LLM(model="mistralai/Pixtral-12B-2409", load_params_config=True)  # Name or path of your model
+    llm = LLM(model="mistralai/Pixtral-12B-2409",
+              tokenizer_mode="mistral",
+              limit_mm_per_prompt={"image": max_img_per_msg})  # Name or path of your model
     messages = [
         {
             "role": "user",