Merge pull request #676 from ShiromiyaG/new-pr

Fix overtrain
IAHispano · Sep 6, 2024 · aced175 · aced175
2 parents 66c4ad9 + 55aefb9
commit aced175
Showing 1 changed file with 5 additions and 3 deletions.
diff --git a/rvc/train/train.py b/rvc/train/train.py
@@ -563,7 +563,7 @@ def train_and_evaluate(
         writers (list): List of TensorBoard writers [writer, writer_eval].
         cache (list): List to cache data in GPU memory.
     """
-    global global_step, lowest_value, loss_disc, consecutive_increases_gen, consecutive_increases_disc
+    global global_step, lowest_value, loss_disc, consecutive_increases_gen, consecutive_increases_disc, smoothed_value_gen, smoothed_value_disc
 
     if epoch == 1:
         lowest_value = {"step": 0, "value": float("inf"), "epoch": 0}
@@ -858,8 +858,10 @@ def train_and_evaluate(
                 ckpt = net_g.module.state_dict()
             else:
                 ckpt = net_g.state_dict()
-            if overtraining_detector != True:
-                overtrain_info = None
+            if overtraining_detector and epoch > 1:
+                overtrain_info = f"Smoothed loss_g {smoothed_value_gen:.3f} and loss_d {smoothed_value_disc:.3f}"
+            else:
+                overtrain_info = ""
             extract_model(
                 ckpt=ckpt,
                 sr=sample_rate,