Corey Morris commited on
Commit
f839734
1 Parent(s): 5d3a9b2

Added new results

Browse files
Files changed (3) hide show
  1. app.py +2 -2
  2. processed_data_2023-10-06.csv +0 -0
  3. results +1 -1
app.py CHANGED
@@ -112,7 +112,7 @@ def find_top_differences_table(df, target_model, closest_models, num_differences
112
 
113
  # st.title('Model Evaluation Results including MMLU by task')
114
  st.title('Interactive Portal for Analyzing Open Source Large Language Models')
115
- st.markdown("""***Last updated September 30th***""")
116
  st.markdown("""**Models that are suspected to have training data contaminated with evaluation data have been removed.**""")
117
  st.markdown("""
118
  Hugging Face runs evaluations on open source models and provides results on a
@@ -124,7 +124,7 @@ st.markdown("""
124
  """)
125
 
126
  # Load the data into memory
127
- data_path = "processed_data_2023-10-05.csv" # Replace with your actual file path
128
  data_df = load_csv_data(data_path)
129
  data_df.rename(columns={"Unnamed: 0": "Model Name"}, inplace=True)
130
  data_df.set_index("Model Name", inplace=True)
 
112
 
113
  # st.title('Model Evaluation Results including MMLU by task')
114
  st.title('Interactive Portal for Analyzing Open Source Large Language Models')
115
+ st.markdown("""***Last updated October 6th***""")
116
  st.markdown("""**Models that are suspected to have training data contaminated with evaluation data have been removed.**""")
117
  st.markdown("""
118
  Hugging Face runs evaluations on open source models and provides results on a
 
124
  """)
125
 
126
  # Load the data into memory
127
+ data_path = "processed_data_2023-10-06.csv"
128
  data_df = load_csv_data(data_path)
129
  data_df.rename(columns={"Unnamed: 0": "Model Name"}, inplace=True)
130
  data_df.set_index("Model Name", inplace=True)
processed_data_2023-10-06.csv ADDED
The diff for this file is too large to render. See raw diff
 
results CHANGED
@@ -1 +1 @@
1
- Subproject commit 738bff578e5d2eea8b09665d1197b91b57c8c1a4
 
1
+ Subproject commit 5a0d74c14c30ce72c2f17ce8f25dfb3645e95d6a