Update README.md

README.md

@@ -193,12 +193,14 @@ model = lwm.from_pretrained(device=device)
 
 ### 9. **Perform Inference**
 
+Before running the inference, it's important to understand the benefits of the different embedding types. The **CLS embeddings (cls_emb)** provide a highly compressed, holistic view of the entire wireless channel, making them ideal for tasks requiring a general understanding, such as classification or high-level decision-making. On the other hand, **channel embeddings (channel_emb)** capture detailed spatial and frequency information from the wireless channel, making them more suitable for complex tasks like beamforming or channel prediction.
+
 You can now perform inference on the preprocessed data using the LWM model.
 
 ```python
 from inference import lwm_inference, create_raw_dataset
 input_types = ['cls_emb', 'channel_emb', 'raw']
-selected_input_type = input_types[0]
+selected_input_type = input_types[1]  # Change the index to select LWM CLS embeddings, LWM channel embeddings, or the original input channels.
 
 if selected_input_type in ['cls_emb', 'channel_emb']:
     dataset = lwm_inference(preprocessed_chs, selected_input_type, model, device)
@@ -206,6 +208,8 @@ else:
     dataset = create_raw_dataset(preprocessed_chs, device)
 ```
 
+By selecting either `cls_emb` or `channel_emb`, you leverage the pre-trained model's rich feature extraction capabilities to transform raw channels into highly informative embeddings. If you prefer to work with the original raw data, you can choose the `raw` input type.
+
 ---
 
 ### 10. **Explore the Interactive Demo**