Update README.md

README.md

@@ -11,4 +11,37 @@ license: apache-2.0
 short_description: A replica of the basic smoltalk chatbot I run locally
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Basic SmolLM2 chatbot:
+
+This is a very basic chatbot using HuggingFaceTB/SmolLM2-[x]-Instruct hosted on the same host as the app. It is basically a replica of the chatbot I have for running locally.
+
+## Hardware Scale Up
+
+- I recommend running this on a little better hardware than this is set up with.
+- I'm using the free tier space, but it needs a few more CPU to write fast enough to be useful.
+- It is rather slow on this setup, but when I run this on my laptop, it works very well on CPU without GPU.
+
+## To run locally
+
+- Download the files or clone the repo.
+- Make sure you have a supported version of transformers and torch installed (or run `pip3 install -r requirements.txt` from the root folder of this repo).
+- Run `python3 app.py` from the root folder of this repo.
+- Set your browser to http://0.0.0.0:7860
+
+
+## Configuration options
+
+- In app.py, there are settings  of the screen.
+    - If you run this locally on a laptop with at least 5CPU cores, I would recommend saving all your local work then set `MODEL` to HuggingFaceTB/SmolLM2-360M-Instruct.
+    - If this works without signs of resource saturation, try setting `MODEL` it to HuggingFaceTB/SmolLM2-1.7B-Instruct. This will write well and works fine on my laptop that is about 2 years old.
+
+```
+MAX_NEW_TOKENS = 250
+MODEL="HuggingFaceTB/SmolLM2-135M-Instruct"
+# MODEL="HuggingFaceTB/SmolLM2-360M-Instruct"
+# MODEL="HuggingFaceTB/SmolLM2-1.7B-Instruct"
+TEMPERATURE = 0.6
+TOP_P = 0.95
+REPETITION_PENALTY = 1.2
+```
+