add qwen3-embedding and all-minilm-l6-v2

tests/test_all_minilm_l6_v2.rs

@@ -0,0 +1,24 @@
+use aha::models::{
+    all_minilm_l6_v2::AllMiniLML6V2Embedding, common::embedding::TextEmbedding,
+};
+use anyhow::Result;
+use std::time::Instant;
+
+#[test]
+fn all_minilm_l6_v2_embedding() -> Result<()> {
+    // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_all_minilm_l6_v2 all_minilm_l6_v2_embedding -r -- --nocapture
+
+    let save_dir =
+        aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    let model_path = format!("{}/sentence-transformers/all-MiniLM-L6-v2/", save_dir);
+
+    let i_start = Instant::now();
+    let mut model = AllMiniLML6V2Embedding::init(&model_path, None, None)?;
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in load model is: {:?}", i_duration);
+    let input_texts = ["test ALL_MINILM_L6_V2 embedding".to_string()];
+    let result = model.embed_texts(&input_texts)?;
+    println!("result: {:?}", result);
+
+    Ok(())
+}

tests/test_qwen3_embedding.rs

@@ -0,0 +1,22 @@
+use aha::models::{common::embedding::TextEmbedding, qwen3_embedding::Qwen3Embedding};
+use anyhow::Result;
+use std::time::Instant;
+
+#[test]
+fn qwen3_embedding() -> Result<()> {
+    // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_qwen3_embedding qwen3_embedding -r -- --nocapture
+
+    let save_dir =
+        aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
+    let model_path = format!("{}/Qwen/Qwen3-Embedding-0.6B/", save_dir);
+
+    let i_start = Instant::now();
+    let mut model = Qwen3Embedding::init(&model_path, None, None)?;
+    let i_duration = i_start.elapsed();
+    println!("Time elapsed in load model is: {:?}", i_duration);
+    let input_texts = ["test Qwen3-Embedding-0.6B".to_string()];
+    let result = model.embed_texts(&input_texts)?;
+    println!("result: {:?}", result);
+
+    Ok(())
+}

tests/weight_test.rs

@@ -2,7 +2,7 @@ use std::collections::HashMap;
 
 use aha::utils::{find_type_files, get_device};
 use anyhow::Result;
-use candle_core::{Device, pickle::read_all_with_key, safetensors};
+use candle_core::{Device, pickle::read_all_with_key, quantized::gguf_file, safetensors};
 use candle_nn::VarBuilder;
 
 #[test]
@@ -310,3 +310,21 @@ fn lfm2vl_weight() -> Result<()> {
     println!("model_list: {:?}", model_list);
     Ok(())
 }
+
+#[test]
+fn gguf_weight() -> Result<()> {
+    // cargo test -F cuda --test weight_test gguf_weight -r -- --nocapture
+    let gguf_path = "/home/jhq/.aha/Qwen/Qwen3-Embedding-0.6B-GGUF/Qwen3-Embedding-0.6B-f16.gguf";
+    let mut model_file = std::fs::File::open(gguf_path)?;
+    let model = gguf_file::Content::read(&mut model_file)?;
+    for (key, value) in model.tensor_infos {
+        println!("{key}: {:#?}", value);
+    }
+    // for (key, value) in model.metadata {
+    //     if key.contains("tokeni") {
+    //         continue;
+    //     }
+    //     println!("{key}: {:#?}", value);
+    // }
+    Ok(())
+}