In python/mxnet/gluon/nn/basic_layers.py:
```
class BatchNorm(HybridBlock):
...
moving_mean_initializer: str or `Initializer`, default 'zeros'
Initializer for the moving mean.
moving_variance_initializer: str or `Initializer`, default 'ones'
Initializer for the moving variance.
...
def __init__(self, axis=1, momentum=0.9, epsilon=1e-5, center=True,
scale=True,
use_global_stats=False, beta_initializer='zeros',
gamma_initializer='ones',
running_mean_initializer='zeros',
running_variance_initializer='ones',
in_channels=0, **kwargs):
```
Should we rename running_mean_initializer to moving_mean_initializer, and
running_variance_initializer to moving_variance_initializer?
Also,
```
self.running_mean = self.params.get('running_mean', grad_req='null',
shape=(in_channels,),
init=running_mean_initializer,
allow_deferred_init=True,
differentiable=False)
self.running_var = self.params.get('running_var', grad_req='null',
shape=(in_channels,),
init=running_variance_initializer,
allow_deferred_init=True,
differentiable=False)
```
'running_mean' and 'running_var' are mismatched with op definition:
```
NNVM_REGISTER_OP(BatchNorm)
.set_attr<nnvm::FListInputNames>("FListInputNames",
[](const NodeAttrs& attrs) {
return std::vector<std::string>{"data", "gamma", "beta", "moving_mean",
"moving_var"};
})
```
This will cause trouble for parameter checking. Shall we align them?
[ Full content available at:
https://github.com/apache/incubator-mxnet/issues/12681 ]
This message was relayed via gitbox.apache.org for [email protected]
