loudnorm,

scripts/exp/train.py

@@ -59,7 +59,7 @@ IGNORE_INDEX = -100
 @argbind.bind("train", "val", without_prefix=True)
 def build_transform():
     transform = transforms.Compose(
-        tfm.VolumeNorm(("uniform", -32, -14)),
+        tfm.VolumeNorm(("uniform", -32, -20)),
         tfm.VolumeChange(("uniform", -6, 3)),
         tfm.RescaleAudio(),
     )

vampnet/modules/base.py

@@ -154,7 +154,7 @@ class VampBase(at.ml.BaseModel):
         start_tokens: Optional[torch.Tensor] = None,
         mask: Optional[torch.Tensor] = None,
         device: str = "cpu",
-        temperature: Union[float, Tuple[float, float]] = 1.0,
+        temperature: Union[float, Tuple[float, float]] = 0.8,
         top_k: int = None,
         sample: str = "gumbel",
         renoise_mode: str = "start",
@@ -267,7 +267,7 @@ class VampBase(at.ml.BaseModel):
         sampling_steps: int = 24,
         start_tokens: Optional[torch.Tensor] = None,
         mask: Optional[torch.Tensor] = None,
-        temperature: Union[float, Tuple[float, float]] = 1.0,
+        temperature: Union[float, Tuple[float, float]] = 0.8,
         top_k: int = None,
         sample: str = "multinomial",
         typical_filtering=False,