update README

evaluation/long_32k_eval/README.md

@@ -0,0 +1,12 @@
+This is our script to evaluate 32k long context task
+
+You need to first install dependencies from requirement.txt
+```
+pip install -r requirement.txt
+```
+
+Then you need to configure the model_path and data_home in *eval_retro_vllm.sh* and then run the following command
+```
+bash run_eval_vllm.sh
+```
+to get the corresponding score