1use anyhow::Result;
2use mistralrs::{IsqType, TextMessageRole, VisionMessages, VisionModelBuilder};
3
4const MODEL_ID: &str = "HuggingFaceTB/SmolVLM-Instruct";
5
6#[tokio::main]
7async fn main() -> Result<()> {
8 let model = VisionModelBuilder::new(MODEL_ID)
9 .with_isq(IsqType::Q8_0)
10 .with_logging()
11 .build()
12 .await?;
13
14 let bytes = match reqwest::blocking::get(
15 "https://cdn.britannica.com/45/5645-050-B9EC0205/head-treasure-flower-disk-flowers-inflorescence-ray.jpg",
16 ) {
17 Ok(http_resp) => http_resp.bytes()?.to_vec(),
18 Err(e) => anyhow::bail!(e),
19 };
20 let image = image::load_from_memory(&bytes)?;
21
22 let messages = VisionMessages::new().add_image_message(
23 TextMessageRole::User,
24 "What is this?",
25 vec![image],
26 &model,
27 )?;
28
29 let response = model.send_chat_request(messages).await?;
30
31 println!("{}", response.choices[0].message.content.as_ref().unwrap());
32 dbg!(
33 response.usage.avg_prompt_tok_per_sec,
34 response.usage.avg_compl_tok_per_sec
35 );
36
37 Ok(())
38}