add LFM2.5-VL-1.6B|LFM2-VL-1.6B

tests/test_lfm2vl.rs

@@ -1,13 +1,71 @@
-use std::time::Instant;
+use std::{pin::pin, time::Instant};
 
-use aha::{chat::ChatCompletionParameters, models::lfm2vl::generate::Lfm2VLGenerateModel};
+use aha::{
+    chat::ChatCompletionParameters,
+    models::{GenerateModel, lfm2vl::generate::Lfm2VLGenerateModel},
+};
 use anyhow::Result;
+use rocket::futures::StreamExt;
 #[test]
 fn lfm2vl_generate() -> Result<()> {
     // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_lfm2vl lfm2vl_generate -r -- --nocapture
-    
+
     let save_dir =
         aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    // let model_path = format!("{}/LiquidAI/LFM2.5-VL-1.6B/", save_dir);
+    let model_path = format!("{}/LiquidAI/LFM2-VL-1.6B/", save_dir);
+    let message = r#"
+    {
+        "model": "lfm2vl",
+        "messages": [
+            {
+                "role": "user",
+                "content": [ 
+                    {
+                        "type": "image",
+                        "image_url": 
+                        {
+                            "url": "file://./assets/img/ocr_test1.png"
+                        }
+                    },             
+                    {
+                        "type": "text", 
+                        "text": "图片里面是什么"
+                    }
+                ]
+            }
+        ]
+    }
+    "#;
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
+    let i_start = Instant::now();
+    let mut model = Lfm2VLGenerateModel::init(&model_path, None, None)?;
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in load model is: {:?}", i_duration);
+
+    let i_start = Instant::now();
+    let result = model.generate(mes)?;
+    let i_duration = i_start.elapsed();
+    println!("generate: \n {:?}", result);
+    if let Some(usage) = &result.usage {
+        let num_token = usage.total_tokens;
+        let duration_secs = i_duration.as_secs_f64();
+        let tps = num_token as f64 / duration_secs;
+        println!("Tokens per second (TPS): {:.2}", tps);
+    }
+    println!("Time elapsed in generate is: {:?}", i_duration);
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn lfm2vl_stream() -> Result<()> {
+    // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_lfm2vl lfm2vl_stream -r -- --nocapture
+    // test with cuda+flash-attn: RUST_BACKTRACE=1 cargo test -F cuda,flash-attn qwen3_0_6b_generate -r -- --nocapture
+
+    let save_dir =
+        aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    // let model_path = format!("{}/LiquidAI/LFM2-1.2B/", save_dir);
     let model_path = format!("{}/LiquidAI/LFM2.5-VL-1.6B/", save_dir);
     let message = r#"
     {
@@ -39,16 +97,13 @@ fn lfm2vl_generate() -> Result<()> {
     println!("Time elapsed in load model is: {:?}", i_duration);
 
     let i_start = Instant::now();
-    let result = model.generate(mes)?;
+    // let result = model.generate(mes)?;
+    let mut stream = pin!(model.generate_stream(mes)?);
     let i_duration = i_start.elapsed();
-    // println!("generate: \n {:?}", result);
-    // if let Some(usage) = &result.usage {
-    //     let num_token = usage.total_tokens;
-    //     let duration_secs = i_duration.as_secs_f64();
-    //     let tps = num_token as f64 / duration_secs;
-    //     println!("Tokens per second (TPS): {:.2}", tps);
-    // }
-    // println!("Time elapsed in generate is: {:?}", i_duration);
+    while let Some(token) = stream.next().await {
+        println!("generate: \n {:?}", token);
+    }
+    println!("Time elapsed in generate is: {:?}", i_duration);
 
     Ok(())
-}
+}