Update README.md

README.md

@@ -22,8 +22,7 @@ inference:
     repetition_penalty: 1.15
     do_sample: true
     temperature: 0.5
-    top_k: 5
-    top_p: 0.9
+    top_p: 0.5
 ---
 
 # ahxt's llama2_xs_460M_experimental trained on the WizardLM's Evol Instruct dataset using AutoTrain
@@ -32,13 +31,20 @@ inference:
 - Dataset: [KnutJaegersberg/WizardLM_evol_instruct_V2_196k_instruct_format](https://huggingface.co/datasets/KnutJaegersberg/WizardLM_evol_instruct_V2_196k_instruct_format)
 - Training: 13.5h under [these parameters](https://huggingface.co/Felladrin/llama2_xs_460M_experimental_evol_instruct/blob/cc151c5669ea37c3ef972e375c74f2d9bfd92b49/training_params.json)
 
-## Prompt format
-
-It was trained in the following prompt format:
+## Recommended Prompt Format
 
 ```
 ### Instruction:
 <instruction>
 
 ### Response:
+```
+
+## Recommended Inference Parameters:
+
+```yml
+repetition_penalty: 1.15
+do_sample: true
+temperature: 0.5
+top_p: 0.5
 ```