formatting

mlcommons · Jul 2, 2024 · e125201 · e125201
1 parent a3e513e
commit e125201
Show file tree

Hide file tree

Showing 2 changed files with 45 additions and 39 deletions.
diff --git a/scoring/performance_profile.py b/scoring/performance_profile.py
@@ -70,28 +70,35 @@
 
 #MPL params
 mpl.rcParams['figure.figsize'] = (16, 10)  # Width, height in inches
-mpl.rcParams['font.family'] = 'serif'      
-mpl.rcParams['font.serif'] = ['Times New Roman'] + mpl.rcParams['font.serif']  # Add Times New Roman as first choice
+mpl.rcParams['font.family'] = 'serif'
+mpl.rcParams['font.serif'] = [
+    'Times New Roman'
+] + mpl.rcParams['font.serif']  # Add Times New Roman as first choice
 mpl.rcParams['font.size'] = 22
 mpl.rcParams['savefig.dpi'] = 300  # Set resolution for saved figures
 
 # Plot Elements
-mpl.rcParams['lines.linewidth'] = 3     # Adjust line thickness if needed
-mpl.rcParams['lines.markersize'] = 6      # Adjust marker size if needed
-mpl.rcParams['axes.prop_cycle'] = mpl.cycler(color=["#1f77b4", "#ff7f0e", "#2ca02c", "#d62728", "#9467bd"])  # Example color cycle (consider ColorBrewer or viridis)
-mpl.rcParams['axes.labelsize'] = 22       # Axis label font size
-mpl.rcParams['xtick.labelsize'] = 20       # Tick label font size
+mpl.rcParams['lines.linewidth'] = 3  # Adjust line thickness if needed
+mpl.rcParams['lines.markersize'] = 6  # Adjust marker size if needed
+mpl.rcParams['axes.prop_cycle'] = mpl.cycler(
+    color=["#1f77b4", "#ff7f0e", "#2ca02c", "#d62728",
+           "#9467bd"])  # Example color cycle (consider ColorBrewer or viridis)
+mpl.rcParams['axes.labelsize'] = 22  # Axis label font size
+mpl.rcParams['xtick.labelsize'] = 20  # Tick label font size
 mpl.rcParams['ytick.labelsize'] = 20
 
 # Legends and Gridlines
-mpl.rcParams['legend.fontsize'] = 20       # Legend font size
-mpl.rcParams['legend.loc'] = 'best'       # Let matplotlib decide the best legend location
-mpl.rcParams['axes.grid'] = True          # Enable grid
-mpl.rcParams['grid.alpha'] = 0.4          # Gridline transparency
+mpl.rcParams['legend.fontsize'] = 20  # Legend font size
+mpl.rcParams[
+    'legend.loc'] = 'best'  # Let matplotlib decide the best legend location
+mpl.rcParams['axes.grid'] = True  # Enable grid
+mpl.rcParams['grid.alpha'] = 0.4  # Gridline transparency
+
 
 def print_dataframe(df):
-      tabulated_df = tabulate(df.T, headers='keys', tablefmt='psql')
-      logging.info(tabulated_df)
+  tabulated_df = tabulate(df.T, headers='keys', tablefmt='psql')
+  logging.info(tabulated_df)
+
 
 def generate_eval_cols(metrics):
   splits = ['train', 'validation']
@@ -206,11 +213,13 @@ def get_workloads_time_to_target(submission,
       num_trials = len(group)
       if num_trials != NUM_TRIALS and not self_tuning_ruleset:
         if strict:
-          raise ValueError(f'In Study {study}: Expecting {NUM_TRIALS} trials for workload '
-                           f'{workload} but found {num_trials} trials.')
+          raise ValueError(
+              f'In Study {study}: Expecting {NUM_TRIALS} trials for workload '
+              f'{workload} but found {num_trials} trials.')
         else:
-          logging.warning(f'In Study {study}: Expecting {NUM_TRIALS} trials for workload '
-                          f'{workload} but found {num_trials} trials.')
+          logging.warning(
+              f'In Study {study}: Expecting {NUM_TRIALS} trials for workload '
+              f'{workload} but found {num_trials} trials.')
 
       # Get trial and time index that reaches target
       trial_idx, time_idx = get_best_trial_index(
@@ -316,9 +325,8 @@ def compute_performance_profiles(submissions,
       # If variants do not have finite score set base_workload score to inf
       base_workload = get_base_workload_name(workload)
       df[base_workload] = df.apply(
-          variant_criteria_filter(base_workload, workload),
-          axis=1)
-
+          variant_criteria_filter(base_workload, workload), axis=1)
+
   logging.info("HELDOUT_WORKLOAD FILTER")
   print_dataframe(df)
 
@@ -415,8 +423,7 @@ def plot_performance_profiles(perf_df,
                               df_col,
                               scale='linear',
                               save_dir=None,
-                              figsize=(30, 10)
-                              ):
+                              figsize=(30, 10)):
   """Plot performance profiles.
 
   Args:
@@ -438,8 +445,7 @@ def plot_performance_profiles(perf_df,
   """
   fig = perf_df.T.plot(figsize=figsize, alpha=0.7)
   df_col_display = f'log10({df_col})' if scale == 'log' else df_col
-  fig.set_xlabel(
-      f'Ratio of `{df_col_display}` to best submission')
+  fig.set_xlabel(f'Ratio of `{df_col_display}` to best submission')
   fig.set_ylabel('Proportion of workloads')
   fig.legend(bbox_to_anchor=(1.0, 1.0))
   plt.tight_layout()

diff --git a/scoring/score_submissions.py b/scoring/score_submissions.py
@@ -124,21 +124,21 @@ def main(_):
   results = {}
   os.makedirs(FLAGS.output_dir, exist_ok=True)
 
-#   for team in os.listdir(FLAGS.submission_directory):
-#     for submission in os.listdir(os.path.join(FLAGS.submission_directory, team)):
-#         print(submission)
-#         experiment_path = os.path.join(FLAGS.submission_directory, team, submission)
-#         df = scoring_utils.get_experiment_df(experiment_path)
-#         results[submission] = df
-#         summary_df = get_submission_summary(df)
-#         with open(os.path.join(FLAGS.output_dir, f'{submission}_summary.csv'),
-#                 'w') as fout:
-#             summary_df.to_csv(fout)
-  
-#   # Save results
-#   with open(os.path.join(FLAGS.output_dir, 'results.pkl'), 'wb') as f:
-#     pickle.dump(results, f)
- 
+  #   for team in os.listdir(FLAGS.submission_directory):
+  #     for submission in os.listdir(os.path.join(FLAGS.submission_directory, team)):
+  #         print(submission)
+  #         experiment_path = os.path.join(FLAGS.submission_directory, team, submission)
+  #         df = scoring_utils.get_experiment_df(experiment_path)
+  #         results[submission] = df
+  #         summary_df = get_submission_summary(df)
+  #         with open(os.path.join(FLAGS.output_dir, f'{submission}_summary.csv'),
+  #                 'w') as fout:
+  #             summary_df.to_csv(fout)
+
+  #   # Save results
+  #   with open(os.path.join(FLAGS.output_dir, 'results.pkl'), 'wb') as f:
+  #     pickle.dump(results, f)
+
   # Read results
   with open(os.path.join(FLAGS.output_dir, 'results.pkl'), 'rb') as f:
     results = pickle.load(f)