Update README.md

README.md

@@ -37,6 +37,12 @@ outputs = model.generate(**inputs, max_new_tokens=200)
 print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
 
+## Using specific checkpoints
+To use a specific checkpoint, use the `revision` argument:
+
+```
+model = AutoModelForCausalLM.from_pretrained("llm360/k2-v2-instruct", device_map="auto", revision="base_0720000")
+```
 ---
 
 ## **Evaluation Summary**