put the model on CPU, because there's no GPU

app.py

@@ -15,7 +15,7 @@ processor = AutoProcessor.from_pretrained("HuggingFaceM4/Idefics3-8B-Llama3")
 model = Idefics3ForConditionalGeneration.from_pretrained("HuggingFaceM4/Idefics3-8B-Llama3", 
         torch_dtype=torch.bfloat16,
         #_attn_implementation="flash_attention_2",
-        trust_remote_code=True).to("cuda")
+        trust_remote_code=True)#.to("cuda")
 
 BAD_WORDS_IDS = processor.tokenizer(["<image>", "<fake_token_around_image>"], add_special_tokens=False).input_ids
 EOS_WORDS_IDS = [processor.tokenizer.eos_token_id]
@@ -50,7 +50,9 @@ def model_inference(
 
     prompt = processor.apply_chat_template(resulting_messages, add_generation_prompt=True)
     inputs = processor(text=prompt, images=[images], return_tensors="pt")
-    inputs = {k: v.to("cuda") for k, v in inputs.items()}
+    # inputs = {k: v.to("cuda") for k, v in inputs.items()}
+    inputs = {k: v for k, v in inputs.items()}
+
 
     generation_args = {
         "max_new_tokens": max_new_tokens,