fix : print train

add : ref image count threshold
change : ray config

image_ref/main.py

@@ -77,7 +77,7 @@ def val_duo(model, data_test, loss_function, epoch, wandb):
     losses = losses / (label.shape[0] * len(data_test.dataset))
     acc = acc / (len(data_test.dataset))
     acc_contrastive = acc_contrastive / (label.shape[0] * len(data_test.dataset))
-    print('Test epoch {}, loss : {:.3f}  acc : {:.3f} acc contrastive : {:.3f}'.format(epoch, losses, acc,
+    print('Test epoch  {}, loss : {:.3f} acc : {:.3f} acc contrastive : {:.3f}'.format(epoch, losses, acc,
                                                                                        acc_contrastive))
 
     if wandb is not None:

image_ref/main_ray.py

@@ -213,9 +213,9 @@ def test_model(best_result, args):
 def main(args, gpus_per_trial=1):
     config = {
         "lr": tune.loguniform(1e-4, 1e-2),
-        "noise": tune.loguniform(1, 1000),
+        "noise": tune.loguniform(1, 10000),
         "p_prop": tune.uniform(5, 95),
-        "optimizer": tune.choice(['Adam', 'SGD']),
+        "optimizer": tune.choice(['Adam', 'SGD']), #adam plus efficace ?
         "sampler": tune.choice(['random', 'balanced']),
     }
     scheduler = ASHAScheduler(

image_ref/utils.py

@@ -173,7 +173,7 @@ if __name__ == '__main__':
     #         print(pep)
     #         f.write(pep+'\n')
     #
-    # df_count = compute_common_peptide("dataset_species_ref_peptides.csv", SPECIES)
+    df_count = compute_common_peptide("dataset_species_ref_peptides.csv", SPECIES)
 
     #
     # Create ref img
@@ -181,15 +181,16 @@ if __name__ == '__main__':
         'fasta/full proteom/steigerwaltii variants/uniparc_proteome_UP000033376_2025_03_14.predicted.parquet')
     min_rt = df_full['RT'].min()
     max_rt = df_full['RT'].max()
-
+    #
     df = pd.read_csv("dataset_species_ref_peptides.csv")
-
+    #
     for spe in SPECIES:
         print(spe)
         df_spe = df[df['Specie'] == spe]
+        df_spec_no_common = df_spe[df_spe['Sequence'].isin(df_count[df_count['Count']<5]['Sequence'])]
         im = build_ref_image_from_diann_global(
-            'fasta/global_peptide_list.parquet', target_seq=df_spe['Sequence'].to_list(), ms1_end_mz=1250,
+            'fasta/global_peptide_list.parquet', target_seq=df_spec_no_common['Sequence'].to_list(), ms1_end_mz=1250,
             ms1_start_mz=350, bin_mz=1, max_cycle=663, min_rt=min_rt, max_rt=max_rt)
         plt.clf()
-        mpimg.imsave('img_ref/' + spe + '.png', im)
-        np.save('img_ref/' + spe + '.npy', im)
+        mpimg.imsave('img_ref_common_th_5/' + spe + '.png', im)
+        np.save('img_ref_common_th_5/' + spe + '.npy', im)