Update README.md

README.md

@@ -18,24 +18,32 @@ This model [giangndm/qwen2.5-omni-7b-mlx](https://huggingface.co/giangndm/qwen2.
 converted to MLX format from [Qwen/Qwen2.5-Omni-7B](https://huggingface.co/Qwen/Qwen2.5-Omni-7B)
 using mlx-lm version **0.24.0**.
 
-## Use with mlx
+## Use with mlx (https://github.com/giangndm/mlx-lm-omni)
 
 ```bash
-pip install mlx-lm
+uv add mlx-lm-omni 
+# or
+uv add https://github.com/giangndm/mlx-lm-omni.git
 ```
 
 ```python
-from mlx_lm import load, generate
+from mlx_lm_omni import load, generate
+import librosa
+from io import BytesIO
+from urllib.request import urlopen
 
-model, tokenizer = load("giangndm/qwen2.5-omni-7b-mlx")
+model, tokenizer = load("giangndm/qwen2.5-omni-7b-mlx-4bit")
 
-prompt = "hello"
+audio_path = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2-Audio/audio/1272-128104-0000.flac"
+audio = librosa.load(BytesIO(urlopen(audio_path).read()), sr=16000)[0]
 
-if tokenizer.chat_template is not None:
-    messages = [{"role": "user", "content": prompt}]
-    prompt = tokenizer.apply_chat_template(
-        messages, add_generation_prompt=True
-    )
+messages = [
+    {"role": "system", "content": "You are a speech recognition model."},
+    {"role": "user", "content": "Transcribe the English audio into text without any punctuation marks.", "audio": audio},
+]
+prompt = tokenizer.apply_chat_template(
+    messages, add_generation_prompt=True
+)
 
-response = generate(model, tokenizer, prompt=prompt, verbose=True)
+text = generate(model, tokenizer, prompt=prompt, verbose=True)
 ```