Corey Morris commited on
Commit
2db58a0
1 Parent(s): 9695a47

Added explanation for the plot and a dataframe of the models

Browse files
Files changed (1) hide show
  1. app.py +7 -1
app.py CHANGED
@@ -259,7 +259,8 @@ else:
259
 
260
 
261
  # Section to select a model and display radar and line charts
262
- st.header("Compare Models")
 
263
  selected_model_name = st.selectbox("Select a Model:", filtered_data.index.tolist())
264
  metrics_to_compare = ['MMLU_abstract_algebra', 'MMLU_astronomy', 'MMLU_business_ethics', 'MMLU_average', 'MMLU_moral_scenarios']
265
  closest_models = filtered_data['MMLU_average'].sub(filtered_data.loc[selected_model_name, 'MMLU_average']).abs().nsmallest(5).index.tolist()
@@ -270,6 +271,11 @@ fig_line = create_line_chart(filtered_data, closest_models, metrics_to_compare)
270
  st.plotly_chart(fig_radar)
271
  st.plotly_chart(fig_line)
272
 
 
 
 
 
 
273
 
274
 
275
  # end of custom scatter plots
 
259
 
260
 
261
  # Section to select a model and display radar and line charts
262
+ st.header("Compare selected models to models the closest 5 models on MMLU average")
263
+ st.write("This is to demonstrate that while the average score is useful, there is a lot of variation in performance on individual tasks.")
264
  selected_model_name = st.selectbox("Select a Model:", filtered_data.index.tolist())
265
  metrics_to_compare = ['MMLU_abstract_algebra', 'MMLU_astronomy', 'MMLU_business_ethics', 'MMLU_average', 'MMLU_moral_scenarios']
266
  closest_models = filtered_data['MMLU_average'].sub(filtered_data.loc[selected_model_name, 'MMLU_average']).abs().nsmallest(5).index.tolist()
 
271
  st.plotly_chart(fig_radar)
272
  st.plotly_chart(fig_line)
273
 
274
+ # show MMLU_average at the beginning of the dataframe
275
+
276
+ st.dataframe(filtered_data.loc[closest_models, metrics_to_compare])
277
+
278
+
279
 
280
 
281
  # end of custom scatter plots