Update app.py

app.py

@@ -37,7 +37,7 @@ kn_max = 2
 
 kns = np.linspace(0.05, 2, 100)
 overheads = []
-for kn in np.linspace(0.2, 2, 100):
+for kn in kns:
     kd = compute_kd(kn)
     overheads.append(compute_overhead(kn, kd)*100)
 
@@ -74,14 +74,15 @@ Optimal model size:\t\t {N_opt/Bn:.2f}B
 Optimal datset size (tokens):\t {D_opt/Bn:.2f}
 
 ## Your setting trade-off:
-Training compute overhead (%):\t {100*compute_overhead(kn, kd):.2f}
+Training compute overhead:\t {100*compute_overhead(kn, kd):.2f}%
 
-Inference cost fraction (%):\t {kn*100:.2f}"""
+Inference cost fraction:\t {kn*100:.2f}%"""
     return text, fig
 
 with gr.Blocks() as demo:
-    N = gr.Number(value=1, label="Model size (in B parameters)")
-    D = gr.Number(value=100, label="Dataset size (in B tokens")
+    gr.Markdown("# Harm's law")
+    N = gr.Number(value=1, label="Model size (in B parameters):")
+    D = gr.Number(value=100, label="Dataset size (in B tokens):")
     button = gr.Button("Compute!")
     
     plot = gr.Plot(value=plt)