File size: 1,500 Bytes
d5175d3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 |
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import torch
import torch.nn as nn
import torch.nn.functional as F
try:
from apex.normalization import FusedLayerNorm as _FusedLayerNorm
has_fused_layernorm = True
class FusedLayerNorm(_FusedLayerNorm):
@torch.jit.unused
def forward(self, x):
if not x.is_cuda:
return super().forward(x)
else:
with torch.cuda.device(x.device):
return super().forward(x)
except ImportError:
has_fused_layernorm = False
def LayerNorm(normalized_shape, eps=1e-5, elementwise_affine=True, export=False):
if torch.jit.is_scripting():
export = True
if not export and torch.cuda.is_available() and has_fused_layernorm:
return FusedLayerNorm(normalized_shape, eps, elementwise_affine)
return torch.nn.LayerNorm(normalized_shape, eps, elementwise_affine)
class Fp32LayerNorm(nn.LayerNorm):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def forward(self, input):
output = F.layer_norm(
input.float(),
self.normalized_shape,
self.weight.float() if self.weight is not None else None,
self.bias.float() if self.bias is not None else None,
self.eps,
)
return output.type_as(input)
|