Make epsilon in BasicNorm learnable, optionally.

egs/librispeech/ASR/conformer_ctc/subsampling.py

@@ -56,7 +56,10 @@ class Conv2dSubsampling(nn.Module):
             DoubleSwish(),
         )
         self.out = ScaledLinear(odim * (((idim - 1) // 2 - 1) // 2), odim)
-        self.out_norm = BasicNorm(odim)
+        # set learn_eps=False because out_norm is preceded by `out`, and `out`
+        # itself has learned scale, so the extra degree of freedom is not
+        # needed.
+        self.out_norm = BasicNorm(odim, learn_eps=False)
         # constrain mean of output to be close to zero.
         self.out_balancer = DerivBalancer(channel_dim=-1, min_positive=0.4, max_positive=0.6)
         self._reset_parameters()
@@ -361,42 +364,45 @@ class BasicNorm(torch.nn.Module):
 
     So the idea is to introduce this large constant value as an explicit
     parameter, that takes the role of the "eps" in LayerNorm, so the network
-    doesn't have to do this trick.
+    doesn't have to do this trick.  We make the "eps" learnable.
-
-    We also introduce a learned scaling factor on the output; and we
-    remove the subtracting-the-mean aspect of LayerNorm (which anyway, is not
-    that useful unless the LayerNorm immediately follows a nonlinearity).
-
 
     Args:
+       num_channels: the number of channels, e.g. 512.
       channel_dim: the axis/dimension corresponding to the channel,
         interprted as an offset from the input's ndim if negative.
         shis is NOT the num_channels; it should typically be one of
         {-2, -1, 0, 1, 2, 3}.
-      initial_eps: the initial "epsilon" that we add as ballast in:
+       eps: the initial "epsilon" that we add as ballast in:
              scale = ((input_vec**2).mean() + epsilon)**-0.5
           Note: our epsilon is actually large, but we keep the name
-        to indicate the connection with normal LayerNorm.
+          to indicate the connection with conventional LayerNorm.
-
+       learn_eps: if true, we learn epsilon; if false, we keep it
-      speed:  a scaling factor that can be interpreted as scaling the learning
+         at the initial value.
-        rate for this module.  CAUTION: the default value of 10.0 intended to be
+      eps_speed: a constant that determines how fast "eps" learns;
-        used with Adam or amsgrad-type optimizers, e.g. Adam or Noam.
+          with Adam and variants, this should probably be >= 1,
-        If you are using SGD you would probably have to set `speed` to
+          e.g. 5.0.  For SGD and variants, probably a value less than one,
-        a value less than one, or the training would be unstable.
+          like 0.1, would be suitable, to prevent instability.
     """
     def __init__(self,
                  num_channels: int,
                  channel_dim: int = -1,  # CAUTION: see documentation.
-                 eps: float = 0.25):
+                 eps: float = 0.25,
+                 learn_eps: bool = True,
+                 eps_speed: float = 5.0):
         super(BasicNorm, self).__init__()
         self.num_channels = num_channels
         self.channel_dim = channel_dim
-        self.eps = eps
+        self.eps_speed = eps_speed
+        if learn_eps:
+            self.eps = nn.Parameter((torch.tensor(eps).log() / eps_speed).detach())
+        else:
+            self.register_buffer('eps', (torch.tensor(eps).log() / eps_speed).detach())
 
 
     def forward(self, x: Tensor) -> Tensor:
         assert x.shape[self.channel_dim] == self.num_channels
-        scales = (torch.mean(x**2, dim=self.channel_dim, keepdim=True) + self.eps) ** -0.5
+        scales = (torch.mean(x**2, dim=self.channel_dim, keepdim=True) +
+                  (self.eps * self.eps_speed).exp()) ** -0.5
         return x * scales
 
 

egs/librispeech/ASR/transducer_stateless/conformer.py

@@ -1129,4 +1129,5 @@ if __name__ == '__main__':
     seq_len = 20
     # Just make sure the forward pass runs.
     f = c(torch.randn(batch_size, seq_len, feature_dim),
-          torch.full((batch_size,), seq_len, dtype=torch.int64))
+          torch.full((batch_size,), seq_len, dtype=torch.int64),
+          warmup_mode=True)

egs/librispeech/ASR/transducer_stateless/train.py

@@ -110,7 +110,7 @@ def get_parser():
     parser.add_argument(
         "--exp-dir",
         type=str,
-        default="transducer_stateless/randcombine1_expscale3_rework2c_maxabs1000_maxp0.95_noexp_convderiv3warmup_embed",
+        default="transducer_stateless/randcombine1_expscale3_rework2c_maxabs1000_maxp0.95_noexp_convderiv3warmup_embed_scale",
         help="""The experiment dir.
         It specifies the directory where all training related
         files, e.g., checkpoints, log, etc, are saved