Update README.md

README.md

@@ -12,6 +12,7 @@ language:
 - es
 - el
 - fr
+- ae
 metrics:
 - bertscore
 base_model:
@@ -32,6 +33,8 @@ tags:
 
 This model is designed to classify audio clips into two categories: "Suno" music or "People" music. It is trained on a dataset containing examples of both types of music and can be used for various applications such as music recommendation, genre classification, and more.
 
+---
+
 ## Model Details
 
 - **Model Name:** `felguk-suno-or-people`
@@ -39,27 +42,44 @@ This model is designed to classify audio clips into two categories: "Suno" music
 - **Input:** Audio clip (WAV format)
 - **Output:** Classification label (`suno` or `people`)
 
+---
+
 ## Usage
 
-You can use this model directly with the Hugging Face `transformers` library. Below is an example of how to load and use the model:
+This model is not currently available via third-party inference providers or the Hugging Face Inference API. However, you can easily use it locally by following the steps below.
 
-```python
-from transformers import pipeline
+### Step 1: Install Required Libraries
 
-# Load the model
-classifier = pipeline("audio-classification", model="Felguk/Felguk-suno-or-people")
+Make sure you have the `transformers` and `datasets` libraries installed:
 
-# Classify an audio file
-result = classifier("path_to_audio_file.wav")
-print(result)
+```bash
+pip install transformers datasets
 ```
-## install
+## load model
 ```bash
-pip install transformers
+from transformers import AutoModelForAudioClassification, AutoFeatureExtractor
+import torch
+
+# Load the model and feature extractor
+model = AutoModelForAudioClassification.from_pretrained("Felguk/Felguk-suno-or-people")
+feature_extractor = AutoFeatureExtractor.from_pretrained("Felguk/Felguk-suno-or-people")
 ```
-#### example
 ```bash
-[
-    {"label": "suno", "score": 0.95},
-    {"label": "people", "score": 0.05}
-]
+from datasets import load_dataset, Audio
+
+# Load an example audio file (replace with your own file)
+dataset = load_dataset("common_voice", "en", split="train", streaming=True)
+audio_sample = next(iter(dataset))["audio"]
+
+# Preprocess the audio
+inputs = feature_extractor(audio_sample["array"], sampling_rate=audio_sample["sampling_rate"], return_tensors="pt")
+```
+```bash
+# Perform inference
+with torch.no_grad():
+    logits = model(**inputs).logits
+
+# Get the predicted label
+predicted_class_id = logits.argmax().item()
+label = model.config.id2label[predicted_class_id]
+print(f"Predicted label: {label}")