add glm-asr-nano

tests/test_glm_asr_nano.rs

@@ -1,11 +1,14 @@
-use aha::models::glm_asr_nano::generate::GlmAsrNanoGenerateModel;
+use std::{pin::pin, time::Instant};
+
+use aha::models::{GenerateModel, glm_asr_nano::generate::GlmAsrNanoGenerateModel};
 use aha_openai_dive::v1::resources::chat::ChatCompletionParameters;
-use anyhow::{Result};
+use anyhow::Result;
+use rocket::futures::StreamExt;
 
 #[test]
 fn glm_asr_nano_generate() -> Result<()> {
     // RUST_BACKTRACE=1 cargo test -F cuda glm_asr_nano_generate -r -- --nocapture
-    let model_path = "/home/jhq/huggingface_model/zai-org/GLM-ASR-Nano-2512/";
+    let model_path = "/home/jhq/huggingface_model/ZhipuAI/GLM-ASR-Nano-2512/";
     let message = r#"
     {
         "model": "glm-asr-nano",
@@ -17,9 +20,9 @@ fn glm_asr_nano_generate() -> Result<()> {
                         "type": "audio",
                         "audio_url": 
                         {
-                            "url": "file://./assets/audio/voice_01.wav"
+                            "url": "https://sis-sample-audio.obs.cn-north-1.myhuaweicloud.com/16k16bit.mp3"
                         }
-                    },              
+                    },           
                     {
                         "type": "text", 
                         "text": "Please transcribe this audio into text"
@@ -30,7 +33,62 @@ fn glm_asr_nano_generate() -> Result<()> {
     }
     "#;
     let mes: ChatCompletionParameters = serde_json::from_str(message)?;
-    let glm_asr_model = GlmAsrNanoGenerateModel::init(model_path, None, None)?;
-    let _ = glm_asr_model.generate(mes)?;
+    let i_start = Instant::now();
+    let mut glm_asr_model = GlmAsrNanoGenerateModel::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 res = glm_asr_model.generate(mes)?;
+    let i_duration = i_start.elapsed();
+    println!("generate: \n {:?}", res);
+    if res.usage.is_some() {
+        let num_token = res.usage.as_ref().unwrap().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 glm_asr_nano_stream() -> Result<()> {
+    // RUST_BACKTRACE=1 cargo test -F cuda glm_asr_nano_stream -r -- --nocapture
+    let model_path = "/home/jhq/huggingface_model/ZhipuAI/GLM-ASR-Nano-2512/";
+    let message = r#"
+    {
+        "model": "glm-asr-nano",
+        "messages": [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "audio",
+                        "audio_url": 
+                        {
+                            "url": "file://./assets/audio/zh.mp3"
+                        }
+                    },           
+                    {
+                        "type": "text", 
+                        "text": "Please transcribe this audio into text"
+                    }
+                ]
+            }
+        ]
+    }
+    "#;
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
+    let i_start = Instant::now();
+    let mut glm_asr_model = GlmAsrNanoGenerateModel::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 mut stream = pin!(glm_asr_model.generate_stream(mes)?);
+    while let Some(item) = stream.next().await {
+        println!("generate: \n {:?}", item);
+    }
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in generate is: {:?}", i_duration);
+    Ok(())
+}