Update app.py

app.py

@@ -85,9 +85,9 @@ Your specificied setting corresponds to the following training compute budget.
 ## Chinchilla optimal:
 If you are optimizeing for model performance and ignore inference cost this is the optimal setting for training:
 
-**Optimal model size: {N_opt/Bn:.2f}B**
+**Optimal model size: {N_opt/Bn:.2f}B parametes**
 
-**Optimal datset size (tokens): {D_opt/Bn:.2f}**
+**Optimal datset size: {D_opt/Bn:.2f}B tokens**
 
 ## Your setting trade-off:
 In both cases 100% corresponds to the compute optimal model.
@@ -100,8 +100,8 @@ In both cases 100% corresponds to the compute optimal model.
 with gr.Blocks() as demo:
     gr.Markdown(INTRO)
     with gr.Row():
-        N = gr.Number(value=1, label="Model size (in B parameters):")
-        D = gr.Number(value=100, label="Dataset size (in B tokens):")
+        N = gr.Number(value=7, label="Model size (in B parameters):")
+        D = gr.Number(value=2000, label="Dataset size (in B tokens):")
     
     button = gr.Button("Compute!")