add MiniCPM5-1B

tests/config_tests.rs

@@ -4,6 +4,7 @@ use aha::models::{
     lfm2::config::Lfm2Config,
     lfm2vl::config::{Lfm2ProcessorConfig, Lfm2VLConfig},
     minicpm4::config::MiniCPM4Config,
+    minicpm5::config::MiniCPM5Config,
     moss_audio_tokenizer_nano::config::MossAudioTokenizerConfig,
     moss_tts_nano::config::MossTTSConfig,
     paddleocr_vl::config::PaddleOCRVLConfig,
@@ -139,3 +140,13 @@ fn moss_tts_config() -> Result<()> {
     println!("{:?}", config);
     Ok(())
 }
+
+#[test]
+fn minicpm5_config() -> Result<()> {
+    // cargo test -F cuda --test config_tests minicpm5_config -r -- --nocapture
+    let model_path = "/home/jhq/.aha/OpenBMB/MiniCPM5-1B/";
+    let config_path = model_path.to_string() + "/config.json";
+    let config: MiniCPM5Config = serde_json::from_slice(&std::fs::read(config_path)?)?;
+    println!("{:?}", config);
+    Ok(())
+}

tests/test_minicpm4.rs

@@ -1,6 +1,6 @@
 use std::{pin::pin, time::Instant};
 
-use aha::models::{GenerateModel, minicpm4::generate::MiniCPMGenerateModel};
+use aha::models::{GenerateModel, minicpm4::generate::MiniCPM4GenerateModel};
 use aha::params::chat::ChatCompletionParameters;
 use anyhow::Result;
 use rocket::futures::StreamExt;
@@ -29,7 +29,7 @@ fn minicpm_generate() -> Result<()> {
     "#;
     let mes: ChatCompletionParameters = serde_json::from_str(message)?;
     let i_start = Instant::now();
-    let mut model = MiniCPMGenerateModel::init(&model_path, None, None)?;
+    let mut model = MiniCPM4GenerateModel::init(&model_path, None, None)?;
     let i_duration = i_start.elapsed();
     println!("Time elapsed in load model is: {:?}", i_duration);
 
@@ -62,7 +62,7 @@ async fn minicpm_stream() -> Result<()> {
     "#;
     let mes: ChatCompletionParameters = serde_json::from_str(message)?;
     let i_start = Instant::now();
-    let mut model = MiniCPMGenerateModel::init(&model_path, None, None)?;
+    let mut model = MiniCPM4GenerateModel::init(&model_path, None, None)?;
     let i_duration = i_start.elapsed();
     println!("Time elapsed in load model is: {:?}", i_duration);
 

tests/test_minicpm5.rs

@@ -0,0 +1,41 @@
+use std::time::Instant;
+
+use aha::{
+    models::{GenerateModel, minicpm5::generate::MiniCPM5GenerateModel},
+    params::chat::ChatCompletionParameters,
+};
+use anyhow::Result;
+#[test]
+fn minicpm5_generate() -> Result<()> {
+    // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_minicpm5 minicpm5_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!("{}/OpenBMB/MiniCPM5-1B/", save_dir);
+    let message = r#"
+    {
+        "temperature": 0.3,
+        "top_p": 0.8,
+        "model": "minicpm4",
+        "messages": [
+            {
+                "role": "user",
+                "content": "什么是AI"
+            }
+        ],
+        "enable_thinking": true
+    }
+    "#;
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
+    let i_start = Instant::now();
+    let mut model = MiniCPM5GenerateModel::init(&model_path, None, None)?;
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in load model is: {:?}", i_duration);
+
+    let res = model.generate(mes)?;
+    println!("generate: \n {:?}", res);
+    if let Some(usage) = &res.usage {
+        println!("usage: \n {:?}", usage);
+    }
+    Ok(())
+}