add halffastresnet

nnet/xsets.py

@@ -85,8 +85,7 @@ class SideSampler(torch.utils.data.Sampler):
         self.num_process = num_process
         self.num_replicas = num_replicas
 
-        assert batch_size % (examples_per_speaker * self.num_replicas) == 0
-        #assert batch_size % examples_per_speaker == 0
+        assert batch_size % examples_per_speaker == 0
         assert (self.samples_per_speaker * self.spk_count * self.examples_per_speaker) % self.num_process == 0
 
         self.batch_size = batch_size // (self.examples_per_speaker * self.num_replicas)

nnet/xvector.py

@@ -995,7 +995,7 @@ def get_network(model_opts, local_rank):
     :return:
     """
 
-    if model_opts["model_type"] in ["xvector", "rawnet2", "resnet34", "fastresnet34"]:
+    if model_opts["model_type"] in ["xvector", "rawnet2", "resnet34", "fastresnet34", "halfresnet34"]:
         model = Xtractor(model_opts["speaker_number"], model_opts["model_type"], loss=model_opts["loss"]["type"])
     else:
         # Custom type of model
@@ -1191,7 +1191,7 @@ def get_optimizer(model, model_opts, train_opts, training_loader):
 
     if train_opts["scheduler"]["type"] == 'CyclicLR':
         cycle_momentum = True
-        if train_opts["optimizer"]["type"] == "aam":
+        if train_opts["optimizer"]["type"] in ["aam", "aps"]:
             cycle_momentum = False
         scheduler = torch.optim.lr_scheduler.CyclicLR(optimizer=optimizer,
                                                       base_lr=1e-8,