data augmented

data/data_viz.py

@@ -136,14 +136,45 @@ def numerical_to_alphabetical_str(s):
         seq+=ALPHABET_UNMOD_REV[arr[i]]
     return seq
 
+def plot_res():
+    import matplotlib.pyplot as plt
+    import numpy as np
+
+    fig, axs = plt.subplots(figsize=(9, 4))
+
+    all_data = [[0.911,0.899,0.9,0.885],
+    [0.852,0.75,0.857,0.788],
+    [0.853,0.839,0.862,0.826],
+    [0.902,0.833,0.808],
+    [0.9,0.904,0.922,0.907],
+    [0.915,0.912,0.923,0.911],
+    [0.945,0.918,0.919,0.933],
+    [0.91,0.919,0.927,0.906],
+    [0.881,0.901,0.919,0.902],
+    [0.893,0.909,0.918,0.896],]
+
+    # plot box plot
+    axs.boxplot(all_data)
+    axs.set_title('Box plot')
+
+    # adding horizontal grid lines
+    axs.yaxis.grid(True)
+    axs.set_xticks([y + 1 for y in range(len(all_data))],
+                  labels=['Prosit','ISA_noc','Augm 0.05','Augm 0.1','Augm 0.2','Augm 0.3','Augm 0.4','Augm 0.7','Augm 1','Augm all',])
+    plt.savefig('../fig/model perf/summary.png')
+
+
+
 if __name__ == '__main__' :
-    base = ['ISA_noc_ISA_noc','prosit_ISA_noc', 'ISA_noc_prosit', 'prosit_prosit']
-    augmented = ['ISA_aug_07_ISA_noc','ISA_aug_1_ISA_noc','ISA_aug_all_ISA_noc']
-    for f_suffix_name in augmented:
-        for number in ['1','2','3','4']:
-            df = pd.read_csv('../output/out_{}_{}.csv'.format(f_suffix_name,number))
-            add_length(df)
-            df['abs_error'] =  np.abs(df['rt pred']-df['true rt'])
+    # base = ['ISA_noc_ISA_noc','prosit_ISA_noc', 'ISA_noc_prosit', 'prosit_prosit']
+    # augmented = ['ISA_aug_07_ISA_noc','ISA_aug_1_ISA_noc','ISA_aug_all_ISA_noc']
+    # for f_suffix_name in augmented:
+    #     for number in ['1','2','3','4']:
+    #         df = pd.read_csv('../output/out_{}_{}.csv'.format(f_suffix_name,number))
+    #         add_length(df)
+    #         df['abs_error'] =  np.abs(df['rt pred']-df['true rt'])
             # histo_abs_error(df, display=False, save=True, path='../fig/model perf/histo_{}_{}.png'.format(f_suffix_name,number))
-            scatter_rt(df, display=False, save=True, path='../fig/model perf/RT_pred_{}_{}.png'.format(f_suffix_name,number), color=True)
-            # histo_length_by_error(df, bins=10, display=False, save=True, path='../fig/model perf/histo_length_{}_{}.png'.format(f_suffix_name,number))
+            # scatter_rt(df, display=False, save=True, path='../fig/model perf/RT_pred_{}_{}.png'.format(f_suffix_name,number), color=True)
+            # histo_length_by_error(df, bins=10, display=False, save=True, path='../fig/model perf/histo_length_{}_{}.png'.format(f_suffix_name,number))
\ No newline at end of file
+    plot_res()
+

main.py

@@ -67,17 +67,24 @@ def eval(model, data_val, epoch, criterion_rt, metric_rt, wandb=None):
     print('epoch : ', epoch, 'val rt loss', losses_rt / len(data_val),
           "val rt mean metric : ",
           dist_rt_acc / len(data_val))
+    return losses_rt
 
 
 def run(epochs, eval_inter, save_inter, model, data_train, data_val, data_test, optimizer, criterion_rt,
          metric_rt, wandb=None, output='output/out.csv'):
+    mem = 1000000.
     for e in range(1, epochs + 1):
         train(model, data_train, e, optimizer, criterion_rt, metric_rt, wandb=wandb)
         if e % eval_inter == 0:
-            eval(model, data_val, e, criterion_rt, metric_rt, wandb=wandb)
+            losses_rt = eval(model, data_val, e, criterion_rt, metric_rt, wandb=wandb)
+            if losses_rt < mem :
+                mem = losses_rt
+                torch.save(model.state_dict(), output.strip('.csv')+'pt')
+                print('model saved')
         if e % save_inter == 0:
             save(model, 'model_common_' + str(e) + '.pt')
-    save_pred(model, data_val, output)
+    model.load_state_dict(torch.load(output.strip('.csv')+'pt', weights_only=True))
+    save_pred(model, data_test, output)
 
 
 def main(args):