Update README.md

README.md

@@ -72,7 +72,7 @@ model = AutoModelForCausalLM.from_pretrained(
 # )
 
 # prepare input
-batch_size, lookback_length = 1, 11520 
+batch_size, lookback_length = 64, 11520 
 seqs = torch.randn(batch_size, lookback_length).to(model.device)
 
 # Note that Timer-S1 generates predictions at fixed quantile levels
@@ -87,18 +87,19 @@ print(output.shape) # batch_size x quantile_num(9) x forecast_length
 print(output[0][4])
 ```
 
-> **Out of GPU memory?** Try the following options:
+
+>  This model support inference using either CPU or GPU. To load this model on GPU, we recommend a GPU with **at least 40GB VRAM** (e.g., A100 40GB/80GB, or H100). **Encounter out-of-memory at runtime?** Try the following options:
 > ```python
 > # Option 1: reduce batch size or context length
 > batch_size, lookback_length = 1, 2880
 >
 > # Option 2: disable KV cache at runtime (or edit it in config.json for a permanent change)
-> model.config.use_cache = False
+> model.config.use_cache = False # there is no efficiency impact for cases where the prediction horizon does not exceed 256.
 > ```
 
 ## Specification
 
-* **Architecture**: decoder-only Transformer
+* **Architecture**: decoder-only Transformer with MoE
 * **Context Length**: up to 11,520
 * **ReNorm**: default=True
 * **KV Cache**: default=True