Update README.md

README.md

@@ -11,9 +11,15 @@ This is a second prototype of SuperHOT, this time with 4K context and no RLHF. I
 - 13B 8K CUDA (no groupsize): [tmpupload/superhot-13b-8k-no-rlhf-test-GPTQ](https://huggingface.co/tmpupload/superhot-13b-8k-no-rlhf-test-GPTQ)
 - 13B 8K CUDA 32g: [tmpupload/superhot-13b-8k-no-rlhf-test-32g-GPTQ](https://huggingface.co/tmpupload/superhot-13b-8k-no-rlhf-test-32g-GPTQ) 
 
-In order to use the 8K context, you will need to apply the monkeypatch I have added in this repo -- **without it, it will not work**.
+#### Using the monkey-patch?
+You will **NEED** to **apply the monkeypatch** or, if you are already using the monkeypatch, **change the scaling factor to 0.25 and the maximum sequence length to 8192**
 
-I will repeat: **Without the patch, it will not work!**
+#### Using Oobabooga with Exllama?
+- `python server.py --max_seq_len 8192 --compress_pos_emb 4 --loader exllama_hf`
+
+In order to use the 8K context, you will need to apply the monkeypatch I have added in this repo or follow the instructions for oobabooga's text-generation-webui -- **without it, it will not work**.
+
+I will repeat: **Without the patch with the correct scaling and max sequence length, it will not work!**
 
 The patch is very simple, and you can make the changes yourself:
 - Increase the `max_position_embeddings` to 8192 to stretch the sinusoidal