Use torch F.rms_norm when possible, select fast vs normal paths appropriately and test with torchscript

Author: rwightman

timm/layers/fast_norm.py

@@ -24,6 +24,8 @@
     has_apex_rmsnorm = False
 
 
+has_torch_rms_norm = hasattr(F, 'rms_norm')
+
 # fast (ie lower precision LN) can be disabled with this flag if issues crop up
 _USE_FAST_NORM = False  # defaulting to False for now
 
@@ -75,7 +77,6 @@ def fast_group_norm(
     if is_autocast_enabled(x.device.type):
         # normally native AMP casts GN inputs to float32
         # here we use the low precision autocast dtype
-        # FIXME what to do re CPU autocast?
         dt = get_autocast_dtype(x.device.type)
         x, weight, bias = x.to(dt), weight.to(dt), bias.to(dt) if bias is not None else None
 
@@ -101,14 +102,12 @@ def fast_layer_norm(
         # normally native AMP casts LN inputs to float32
         # apex LN does not, this is behaving like Apex
         dt = get_autocast_dtype(x.device.type)
-        # FIXME what to do re CPU autocast?
         x, weight, bias = x.to(dt), weight.to(dt), bias.to(dt) if bias is not None else None
 
     with torch.amp.autocast(device_type=x.device.type, enabled=False):
         return F.layer_norm(x, normalized_shape, weight, bias, eps)
 
 
-
 def rms_norm(
     x: torch.Tensor,
     normalized_shape: List[int],
@@ -148,8 +147,19 @@ def fast_rms_norm(
         else:
             return fused_rms_norm_affine(x, weight, normalized_shape, eps)
 
-    # fallback
-    return rms_norm(x, normalized_shape, weight, eps)
+    if is_autocast_enabled(x.device.type):
+        # normally native AMP casts LN inputs to float32
+        # apex LN does not, this is behaving like Apex
+        dt = get_autocast_dtype(x.device.type)
+        x, weight = x.to(dt), weight.to(dt)
+
+    with torch.amp.autocast(device_type=x.device.type, enabled=False):
+        if has_torch_rms_norm:
+            x = F.rms_norm(x, normalized_shape, weight, eps)
+        else:
+            x = rms_norm(x, normalized_shape, weight, eps)
+
+    return x
 
 
 def simple_norm(

timm/layers/norm.py

@@ -11,17 +11,24 @@
 import torch.nn as nn
 import torch.nn.functional as F
 
-from .fast_norm import is_fast_norm, fast_group_norm, fast_layer_norm, fast_rms_norm, fast_simple_norm
+from .fast_norm import is_fast_norm, fast_group_norm, fast_layer_norm, fast_rms_norm, fast_simple_norm, simple_norm
+
+try:
+    from torch.nn.functional import rms_norm
+except ImportError:
+    from .fast_norm import rms_norm
 
 
 class GroupNorm(nn.GroupNorm):
+    _fast_norm: torch.jit.Final[bool]
+
     def __init__(self, num_channels, num_groups=32, eps=1e-5, affine=True):
         # NOTE num_channels is swapped to first arg for consistency in swapping norm layers with BN
         super().__init__(num_groups, num_channels, eps=eps, affine=affine)
-        self.fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
 
     def forward(self, x):
-        if self.fast_norm:
+        if self._fast_norm:
             return fast_group_norm(x, self.num_groups, self.weight, self.bias, self.eps)
         else:
             return F.group_norm(x, self.num_groups, self.weight, self.bias, self.eps)
@@ -31,13 +38,14 @@ class GroupNorm1(nn.GroupNorm):
     """ Group Normalization with 1 group.
     Input: tensor in shape [B, C, *]
     """
+    _fast_norm: torch.jit.Final[bool]
 
     def __init__(self, num_channels, **kwargs):
         super().__init__(1, num_channels, **kwargs)
-        self.fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
-        if self.fast_norm:
+        if self._fast_norm:
             return fast_group_norm(x, self.num_groups, self.weight, self.bias, self.eps)
         else:
             return F.group_norm(x, self.num_groups, self.weight, self.bias, self.eps)
@@ -46,6 +54,8 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 class LayerNorm(nn.LayerNorm):
     """ LayerNorm w/ fast norm option
     """
+    _fast_norm: torch.jit.Final[bool]
+
     def __init__(self, num_channels, eps=1e-6, affine=True):
         super().__init__(num_channels, eps=eps, elementwise_affine=affine)
         self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
@@ -60,6 +70,8 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 
 class LayerNorm2d(nn.LayerNorm):
     """ LayerNorm for channels of '2D' spatial NCHW tensors """
+    _fast_norm: torch.jit.Final[bool]
+
     def __init__(self, num_channels, eps=1e-6, affine=True):
         super().__init__(num_channels, eps=eps, elementwise_affine=affine)
         self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
@@ -121,10 +133,11 @@ def forward(self, x) -> torch.Tensor:
 class RmsNorm(nn.Module):
     """ RmsNorm w/ fast (apex) norm if available
     """
-    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine']
+    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine', '_fast_norm']
     normalized_shape: Tuple[int, ...]
     eps: float
     elementwise_affine: bool
+    _fast_norm: bool
 
     def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) -> None:
         factory_kwargs = {'device': device, 'dtype': dtype}
@@ -136,6 +149,8 @@ def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) ->
         self.normalized_shape = tuple(normalized_shape)  # type: ignore[arg-type]
         self.eps = eps
         self.elementwise_affine = affine
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+
         if self.elementwise_affine:
             self.weight = nn.Parameter(torch.empty(self.normalized_shape, **factory_kwargs))
         else:
@@ -150,17 +165,21 @@ def reset_parameters(self) -> None:
     def forward(self, x: torch.Tensor) -> torch.Tensor:
         # NOTE fast norm fallback needs our rms norm impl, so both paths through here.
         # Since there is no built-in PyTorch impl, always use APEX RmsNorm if is installed.
-        x = fast_rms_norm(x, self.normalized_shape, self.weight, self.eps)
+        if self._fast_norm:
+            x = fast_rms_norm(x, self.normalized_shape, self.weight, self.eps)
+        else:
+            x = rms_norm(x, self.normalized_shape, self.weight, self.eps)
         return x
 
 
 class RmsNorm2d(nn.Module):
     """ RmsNorm w/ fast (apex) norm if available
     """
-    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine']
+    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine', '_fast_norm']
     normalized_shape: Tuple[int, ...]
     eps: float
     elementwise_affine: bool
+    _fast_norm: bool
 
     def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) -> None:
         factory_kwargs = {'device': device, 'dtype': dtype}
@@ -172,6 +191,8 @@ def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) ->
         self.normalized_shape = tuple(normalized_shape)  # type: ignore[arg-type]
         self.eps = eps
         self.elementwise_affine = affine
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+
         if self.elementwise_affine:
             self.weight = nn.Parameter(torch.empty(self.normalized_shape, **factory_kwargs))
         else:
@@ -187,18 +208,22 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         x = x.permute(0, 2, 3, 1)
         # NOTE fast norm fallback needs our rms norm impl, so both paths through here.
         # Since there is no built-in PyTorch impl, always use APEX RmsNorm if is installed.
-        x = fast_rms_norm(x, self.normalized_shape, self.weight, self.eps)
+        if self._fast_norm:
+            x = fast_rms_norm(x, self.normalized_shape, self.weight, self.eps)
+        else:
+            x = rms_norm(x, self.normalized_shape, self.weight, self.eps)
         x = x.permute(0, 3, 1, 2)
         return x
 
 
 class SimpleNorm(nn.Module):
     """ SimpleNorm (x / std(x))
     """
-    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine']
+    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine', '_fast_norm']
     normalized_shape: Tuple[int, ...]
     eps: float
     elementwise_affine: bool
+    _fast_norm: bool
 
     def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) -> None:
         factory_kwargs = {'device': device, 'dtype': dtype}
@@ -210,6 +235,8 @@ def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) ->
         self.normalized_shape = tuple(normalized_shape)  # type: ignore[arg-type]
         self.eps = eps
         self.elementwise_affine = affine
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+
         if self.elementwise_affine:
             self.weight = nn.Parameter(torch.empty(self.normalized_shape, **factory_kwargs))
         else:
@@ -222,17 +249,21 @@ def reset_parameters(self) -> None:
             nn.init.ones_(self.weight)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
-        x = fast_simple_norm(x, self.normalized_shape, self.weight, self.eps)
+        if self._fast_norm:
+            x = fast_simple_norm(x, self.normalized_shape, self.weight, self.eps)
+        else:
+            x = simple_norm(x, self.normalized_shape, self.weight, self.eps)
         return x
 
 
 class SimpleNorm2d(nn.Module):
     """ SimpleNorm for NCHW tensors
     """
-    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine']
+    __constants__ = ['normalized_shape', 'eps', 'elementwise_affine', '_fast_norm']
     normalized_shape: Tuple[int, ...]
     eps: float
     elementwise_affine: bool
+    _fast_norm: bool
 
     def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) -> None:
         factory_kwargs = {'device': device, 'dtype': dtype}
@@ -244,6 +275,8 @@ def __init__(self, channels, eps=1e-6, affine=True, device=None, dtype=None) ->
         self.normalized_shape = tuple(normalized_shape)  # type: ignore[arg-type]
         self.eps = eps
         self.elementwise_affine = affine
+        self._fast_norm = is_fast_norm()  # can't script unless we have these flags here (no globals)
+
         if self.elementwise_affine:
             self.weight = nn.Parameter(torch.empty(self.normalized_shape, **factory_kwargs))
         else:
@@ -257,6 +290,9 @@ def reset_parameters(self) -> None:
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
         x = x.permute(0, 2, 3, 1)
-        x = fast_simple_norm(x, self.normalized_shape, self.weight, self.eps)
+        if self._fast_norm:
+            x = fast_simple_norm(x, self.normalized_shape, self.weight, self.eps)
+        else:
+            x = simple_norm(x, self.normalized_shape, self.weight, self.eps)
         x = x.permute(0, 3, 1, 2)
         return x