add Pixtral

app.py

@@ -38,23 +38,11 @@ with open(f'{model_path}/params.json', 'r') as f:
 with open(f'{model_path}/tekken.json', 'r') as f:
     tokenizer_config = json.load(f)
 
-@spaces.GPU()
-def initialize_llm():
-    try:
-        llm = LLM(
-            model=repo_id,
-            tokenizer_mode="mistral",
-            max_model_len=65536,
-            max_num_batched_tokens=max_img_per_msg * max_tokens_per_img,
-            limit_mm_per_prompt={"image": max_img_per_msg}
-        )
-        return llm
-    except Exception as e:
-        print("LLM initialization failed:", e)
-        return None
+model_name = "mistralai/Pixtral-12B-2409"
 
 sampling_params = SamplingParams(max_tokens=8192)
-llm = initialize_llm()
+
+llm = LLM(model=model_name, tokenizer_mode="mistral")
 
 def encode_image(image: Image.Image, image_format="PNG") -> str:
     im_file = BytesIO()
@@ -63,7 +51,6 @@ def encode_image(image: Image.Image, image_format="PNG") -> str:
     im_64 = base64.b64encode(im_bytes).decode("utf-8")
     return im_64
 
-@spaces.GPU()
 def infer(image_url, prompt, progress=gr.Progress(track_tqdm=True)):
     if llm is None:
         return "Error: LLM initialization failed. Please try again later."
@@ -86,7 +73,6 @@ def infer(image_url, prompt, progress=gr.Progress(track_tqdm=True)):
     except Exception as e:
         return f"Error during inference: {e}"
 
-@spaces.GPU()
 def compare_images(image1_url, image2_url, prompt, progress=gr.Progress(track_tqdm=True)):
     if llm is None:
         return "Error: LLM initialization failed. Please try again later."
@@ -116,7 +102,6 @@ def compare_images(image1_url, image2_url, prompt, progress=gr.Progress(track_tq
     except Exception as e:
         return f"Error during image comparison: {e}"
 
-@spaces.GPU()
 def calculate_image_similarity(image1_url, image2_url):
     if llm is None:
         return "Error: LLM initialization failed. Please try again later."