add lfm2vl

tests/config_tests.rs

@@ -1,8 +1,5 @@
 use aha::models::{
-    deepseek_ocr::config::DeepseekOCRConfig, hunyuan_ocr::config::HunYuanVLConfig,
-    lfm2::config::Lfm2Config, minicpm4::config::MiniCPM4Config,
-    paddleocr_vl::config::PaddleOCRVLConfig, qwen2_5vl::config::Qwen2_5VLConfig,
-    qwen3vl::config::Qwen3VLConfig, voxcpm::config::VoxCPMConfig,
+    deepseek_ocr::config::DeepseekOCRConfig, hunyuan_ocr::config::HunYuanVLConfig, lfm2::config::Lfm2Config, lfm2vl::config::{Lfm2ProcessorConfig, Lfm2VLConfig}, minicpm4::config::MiniCPM4Config, paddleocr_vl::config::PaddleOCRVLConfig, qwen2_5vl::config::Qwen2_5VLConfig, qwen3vl::config::Qwen3VLConfig, voxcpm::config::VoxCPMConfig
 };
 use anyhow::Result;
 
@@ -98,3 +95,16 @@ fn lfm2_config() -> Result<()> {
     println!("{:?}", config);
     Ok(())
 }
+
+#[test]
+fn lfm2vl_config() -> Result<()> {
+    // cargo test -F cuda --test config_tests lfm2vl_config -r -- --nocapture
+    let model_path = "/home/jhq/.aha/LiquidAI/LFM2.5-VL-1.6B/";
+    let config_path = model_path.to_string() + "/config.json";
+    let config: Lfm2VLConfig = serde_json::from_slice(&std::fs::read(config_path)?)?;
+    println!("{:?}", config);
+    let processor_config_path = model_path.to_string() + "/processor_config.json";
+    let processor_config: Lfm2ProcessorConfig = serde_json::from_slice(&std::fs::read(processor_config_path)?)?;
+    println!("{:?}", processor_config);
+    Ok(())
+}

tests/test_lfm2vl.rs

@@ -0,0 +1,54 @@
+use std::time::Instant;
+
+use aha::{chat::ChatCompletionParameters, models::lfm2vl::generate::Lfm2VLGenerateModel};
+use anyhow::Result;
+#[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 message = r#"
+    {
+        "model": "lfm2vl",
+        "messages": [
+            {
+                "role": "user",
+                "content": [ 
+                    {
+                        "type": "image",
+                        "image_url": 
+                        {
+                            "url": "file://./assets/img/ocr_test1.png"
+                        }
+                    },             
+                    {
+                        "type": "text", 
+                        "text": "请分析图片并提取所有可见文本内容，按从左到右、从上到下的布局，返回纯文本"
+                    }
+                ]
+            }
+        ]
+    }
+    "#;
+    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(())
+}