You are viewing a plain text version of this content. The canonical link for it is here.
Posted to discuss-archive@mxnet.apache.org by Antonie Lin via MXNet Forum <mx...@discoursemail.com.INVALID> on 2021/04/08 02:44:01 UTC
[MXNet Forum] [Gluon] How to apply F.softmax to Gluon Parameters
I am trying to implement [task-specific weighting of multiple embeddings as in Elmo](https://arxiv.org/pdf/1802.05365.pdf).
Currently, I initialized weights for multiple embeddings using `self.param.get`. However, it throws me the error.
`AssertionError: Argument data must be Symbol instances, but got Parameter elmoembedding0_weights (shape=(3,), dtype=<class 'numpy.float32'>)`.
I can call `x.data()` for non hybridized or `x.var()` for hybridized version for the parameters. Is there a way to simply apply Softmax to parameters and work with both versions? Thanks!
My code looks something like this
```
import mxnet as mx
import mxnet.gluon as gluon
class ElmoEmbedding(gluon.HybridBlock):
def __init__(self):
super(ElmoEmbedding, self).__init__()
with self.name_scope():
self.weights = self.params.get('weights',
shape=(3,),
init=mx.init.Constant(1.0))
self.scales = self.params.get('scales',
shape=(1,0),
init=mx.init.Constant(1.0))
def hybrid_forward(self, F, x, *args, **kwargs):
normalized_weights = F.softmax(self.weights)
weighted_x = F.dot(normalized_weights, x)
output = F.broadcast_mul(self.scales, weighted_x)
return output
net = ElmoEmbedding()
net.hybridize()
# create input
x = mx.ndarray.random.randn(3,100)
output = net(x)
print("output", output.shape)
```
---
[Visit Topic](https://discuss.mxnet.apache.org/t/how-to-apply-f-softmax-to-gluon-parameters/6930/1) or reply to this email to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.mxnet.apache.org/email/unsubscribe/97036f3ac6bc46f8d999282ce625d054f6fcc6ddb3d52b75e7e615d16c3d287d).
[MXNet Forum] [Gluon] How to apply F.softmax to Gluon Parameters
Posted by Amir Ramezani via MXNet Forum <mx...@discoursemail.com.INVALID>.
hi,
initialize your network before calling net.hybridize() and see what happens
---
[Visit Topic](https://discuss.mxnet.apache.org/t/how-to-apply-f-softmax-to-gluon-parameters/6930/2) or reply to this email to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.mxnet.apache.org/email/unsubscribe/538b94daa129343f10fe0f063caec1190eeb91b7efeafd8732f1d4ab15fe0c55).
[MXNet Forum] [Gluon] How to apply F.softmax to Gluon Parameters
Posted by Antonie Lin via MXNet Forum <mx...@discoursemail.com.INVALID>.
`hybrid_forward` passes the parameters as part of the function arguments, so you would access them by defining them or using `kwargs`. Then we can use them as usual for both hybridized and non-hybridized versions.
```
class ElmoEmbedding(gluon.HybridBlock):
def __init__(self):
...
def hybrid_forward(self, F, x, *args, **kwargs):
weights = kwargs['weights']
scales = kwargs['scales']
normalized_weights = F.softmax(weights)
...
```
---
[Visit Topic](https://discuss.mxnet.apache.org/t/how-to-apply-f-softmax-to-gluon-parameters/6930/3) or reply to this email to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.mxnet.apache.org/email/unsubscribe/53ea94e87ad8805344f741777f41d4c406ca72079ba1973e7dbbd93a433b7ce7).