merge glm-ocr model

2026-03-07 18:08:06 +08:00
parent cf308f93eb
commit d2524d6a19
19 changed files with 395 additions and 775 deletions
@@ -0,0 +1,99 @@
+use std::{pin::pin, time::Instant};
+
+use aha::models::{GenerateModel, glm_ocr::generate::GlmOcrGenerateModel};
+use aha_openai_dive::v1::resources::chat::ChatCompletionParameters;
+use anyhow::Result;
+use rocket::futures::StreamExt;
+
+#[test]
+fn glm_ocr_generate() -> Result<()> {
+    // RUST_BACKTRACE=1 cargo test -F cuda --test test_glm_ocr glm_ocr_generate -r -- --nocapture
+    let message = r#"
+    {
+        "model": "glm-ocr",
+        "messages": [
+            {
+                "role": "user",
+                "content": [ 
+                    {
+                        "type": "image",
+                        "image_url": 
+                        {
+                            "url": "file://./assets/img/ocr_test1.png"
+                        }
+                    },              
+                    {
+                        "type": "text", 
+                        "text": "Text Recognition:"
+                    }
+                ]
+            }
+        ]
+    }
+    "#;
+    let save_dir =
+        aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    let model_path = format!("{}/ZhipuAI/GLM-OCR/", save_dir);
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
+    let i_start = Instant::now();
+    let mut model = GlmOcrGenerateModel::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 = 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_ocr_stream() -> Result<()> {
+    // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda glm_ocr_stream -r -- --nocapture
+
+    let message = r#"
+    {
+        "model": "glm-ocr",
+        "messages": [
+            {
+                "role": "user",
+                "content": [ 
+                    {
+                        "type": "image",
+                        "image_url": 
+                        {
+                            "url": "file://./assets/img/ocr_test1.png"
+                        }
+                    },              
+                    {
+                        "type": "text", 
+                        "text": "Text Recognition:"
+                    }
+                ]
+            }
+        ]
+    }
+    "#;
+    let save_dir =
+        aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    let model_path = format!("{}/ZhipuAI/GLM-OCR/", save_dir);
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
+    let i_start = Instant::now();
+    let mut model = GlmOcrGenerateModel::init(&model_path, None, None)?;
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in load model is: {:?}", i_duration);
+    let mut stream = pin!(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(())
+}