mirrors/pytorch-image-models
1
0
Fork 0
mirror of https://github.com/huggingface/pytorch-image-models.git synced 2025-06-03 15:01:08 +08:00
Code Issues Projects Releases Wiki Activity

Enable fixed fanout calc in EfficientNet/MobileNetV3 weight init by default. Fix #84

Browse Source
This commit is contained in:
Ross Wightman 2020-02-24 15:11:26 -08:00
parent 27b3680d49
commit 9fee316752
1 changed files with 2 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
timm/models/efficientnet_builder.py
View File

@ -359,15 +359,13 @@ class EfficientNetBuilder:
return stages
def _init_weight_goog(m, n='', fix_group_fanout=False):
def _init_weight_goog(m, n='', fix_group_fanout=True):
""" Weight initialization as per Tensorflow official implementations.
Args:
m (nn.Module): module to init
n (str): module name
fix_group_fanout (bool): enable correct fanout calculation w/ group convs
FIXME change fix_group_fanout to default to True if experiments show better training results
fix_group_fanout (bool): enable correct (matching Tensorflow TPU impl) fanout calculation w/ group convs
Handles layers in EfficientNet, EfficientNet-CondConv, MixNet, MnasNet, MobileNetV3, etc:
* https://github.com/tensorflow/tpu/blob/master/models/official/mnasnet/mnasnet_model.py