debug

lium_baseline/system.py

@@ -228,6 +228,7 @@ def extract_vectors(current_diar, root_folder, model_cfg, show, model=None):
 
     elif model_cfg["model"]["type"] == "lium_xv":
 
+
         current_im = current_diar.id_map()
         # current_im.write(f"{file_path}/{show}.idmap.h5")
 
@@ -242,9 +243,31 @@ def extract_vectors(current_diar, root_folder, model_cfg, show, model=None):
                                                                           sample_rate=16000,
                                                                           mixed_precision=False)
 
-        current_vec = current_vec_per_segment.mean_stat_per_model()
-
-    return current_vec, current_vec_per_segment
+        current_vec_per_cluster = current_vec_per_segment.mean_stat_per_model()
+        #current_vec_per_cluster.norm_stat1()
+        #current_vec_per_cluster= sidekit.nnet.xvector.extract_embeddings_per_speaker(idmap_name=current_im,
+        #                                                                             model_filename=xtractor_name,
+        #                                                                             data_root_name=f"{root_folder}/wav/",
+        #                                                                             device=torch.device("cuda"),
+        #                                                                             num_thread=5)
+        print(f"nombre de segments: {current_vec_per_segment.stat1.shape}, nombre de locuteurs: {current_vec_per_cluster.stat1.shape}")
+        #diar_seg=copy.deepcopy(diar)
+        #for i in range(len(diar_seg)):
+        #    diar_seg[i]["cluster"]="tmp_"+str(i)
+        #current_im = diar_seg.id_map()
+        #current_im.start = current_im.start * 160
+        #current_im.stop = current_im.stop * 160
+        #if os.path.exists(f"{file_path}/{out_file_name}_seg.idmap.h5"):
+        #    os.remove(f"{file_path}/{out_file_name}_seg.idmap.h5")
+        #current_im.write(f"{file_path}/{out_file_name}_seg.idmap.h5")
+        #current_vec_per_seg= sidekit.nnet.xvector.extract_embeddings_per_speaker(idmap_name=f"{file_path}/{out_file_name}_seg.idmap.h5",
+        #                                                                         model_filename=f"{model_cfg['model_dir']}/best_xtractor.pt",
+        #                                                                         data_root_name=f"{model_cfg['wav_dir']}",
+        #                                                                         device=torch.device("cuda"),
+        #                                                                         transform_pipeline=model_cfg["model"]["vectors"]["xvectors"]["transforms"],
+        #                                                                         num_thread=5)
+
+    return current_vec_per_cluster, current_vec_per_segment
 
 
 def perform_second_seg(model,