Update README.md

README.md

@@ -16,6 +16,12 @@ tags:
 
 ![](https://huggingface.co/JosephusCheung/Guanaco/resolve/main/StupidBanner.png)
 
+**You need a Colab pro for full version (8bit not well performed, should use fp16)** 
+
+[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1ocSmoy3ba1EkYu7JWT1oCw9vz8qC2cMk#scrollTo=zLORi5OcPcIJ)
+
+Free T4 Colab demo, please check 4bit version [JosephusCheung/GuanacoOnConsumerHardware](https://huggingface.co/JosephusCheung/GuanacoOnConsumerHardware).**
+
  **It is highly recommended to use fp16 inference for this model, as 8-bit precision may significantly affect performance. If you require a more Consumer Hardware friendly version, please use the specialized quantized [JosephusCheung/GuanacoOnConsumerHardware](https://huggingface.co/JosephusCheung/GuanacoOnConsumerHardware).**
 
  **You are encouraged to use the latest version of transformers from GitHub.**