diff --git a/LOCAL.md b/LOCAL.md
index 931b1bf..3008ebe 100644
--- a/LOCAL.md
+++ b/LOCAL.md
@@ -96,6 +96,7 @@
 11. Start Training.  
    Run `python finetune_speaker_v2.py -m "./OUTPUT_MODEL" --max_epochs "{Maximum_epochs}" --drop_speaker_embed True`  
    Do replace `{Maximum_epochs}` with your desired number of epochs. Empirically, 100 or more is recommended.  
+   To continue training on previous checkpoint, change the training command to: `python finetune_speaker_v2.py -m "./OUTPUT_MODEL" --max_epochs "{Maximum_epochs}" --drop_speaker_embed True --cont True`. Before you do this, make sure you have previous `G_latest.pth` and `D_latest.pth` under `./OUTPUT_MODEL/` directory.  
    To view training progress, open a new terminal and `cd` to the project root directory, run `tensorboard --logdir="./OUTPUT_MODEL"`, then visit `localhost:6006` with your web browser.
 12. After training is completed, you can use your model by running:  
    `python VC_inference.py --model_dir ./OUTPUT_MODEL/G_latest.pth --share True`
diff --git a/finetune_speaker_v2.py b/finetune_speaker_v2.py
index 85fa044..809000e 100644
--- a/finetune_speaker_v2.py
+++ b/finetune_speaker_v2.py
@@ -98,8 +98,17 @@ def run(rank, n_gpus, hps):
   net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(rank)
 
   # load existing model
-  _, _, _, _ = utils.load_checkpoint("./pretrained_models/G_0.pth", net_g, None, drop_speaker_emb=hps.drop_speaker_embed)
-  _, _, _, _ = utils.load_checkpoint("./pretrained_models/D_0.pth", net_d, None)
+  G_ckpt = "./pretrained_models/G_latest.pth" if hps.cont else "./pretrained_models/G_0.pth"
+  D_ckpt = "./pretrained_models/D_latest.pth" if hps.cont else "./pretrained_models/D_0.pth"
+  try:
+      _, _, _, _ = utils.load_checkpoint(G_ckpt, net_g, None,
+                                         drop_speaker_emb=hps.drop_speaker_embed)
+  except Exception:
+      _, _, _, _ = utils.load_checkpoint("./pretrained_models/G_0.pth", net_g, None, drop_speaker_emb=hps.drop_speaker_embed)
+  try:
+      _, _, _, _ = utils.load_checkpoint(D_ckpt, net_d, None)
+  except Exception:
+      _, _, _, _ = utils.load_checkpoint("./pretrained_models/D_0.pth", net_d, None)
   epoch_str = 1
   global_step = 0
   # freeze all other layers except speaker embedding
@@ -243,13 +252,15 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade
         utils.save_checkpoint(net_g, None, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "G_{}.pth".format(global_step)))
         utils.save_checkpoint(net_g, None, hps.train.learning_rate, epoch,
                               os.path.join(hps.model_dir, "G_latest.pth".format(global_step)))
-        # utils.save_checkpoint(net_d, optim_d, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "D_{}.pth".format(global_step)))
+        utils.save_checkpoint(net_d, None, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "D_{}.pth".format(global_step)))
+        utils.save_checkpoint(net_d, None, hps.train.learning_rate, epoch,
+                              os.path.join(hps.model_dir, "D_latest.pth".format(global_step)))
         old_g=os.path.join(hps.model_dir, "G_{}.pth".format(global_step-4000))
-        # old_d=os.path.join(hps.model_dir, "D_{}.pth".format(global_step-400))
+        old_d=os.path.join(hps.model_dir, "D_{}.pth".format(global_step-4000))
         if os.path.exists(old_g):
           os.remove(old_g)
-        # if os.path.exists(old_d):
-        #   os.remove(old_d)
+        if os.path.exists(old_d):
+          os.remove(old_d)
     global_step += 1
     if epoch > hps.max_epochs:
         print("Maximum epoch reached, closing training...")
diff --git a/utils.py b/utils.py
index d72ea0f..954db6a 100644
--- a/utils.py
+++ b/utils.py
@@ -286,6 +286,7 @@ def get_hparams(init=True):
                         help='Model name')
     parser.add_argument('-n', '--max_epochs', type=int, default=50,
                         help='finetune epochs')
+    parser.add_argument('--cont', type=bool, default=False, help='whether to continue training on the latest checkpoint')
     parser.add_argument('--drop_speaker_embed', type=bool, default=False, help='whether to drop existing characters')
 
     args = parser.parse_args()