-
It seems like our loss is jumping to significant values only for some particular instances. We do not have any proper normalization of our data right now, so I think that might be the problem.
-
Construct different mini-batches for real and fake, i.e. each mini-batch needs to contain only all real images or all generated images.
when batchnorm is not an option use instance normalization (for…
-
test_operator.test_norm causes hangs at Python 3: GPU Win.
http://jenkins.mxnet-ci.amazon-ml.com/view/Master%20Jobs/job/incubator-mxnet/job/master/1109/
```
test_operator.test_pad ... ok
te…
-
Hello,
Thank you for the amazing work and for providing access to this Github repo. I have a quick question. After generating the normalized matrices using all 9 different normalization techniques…
-
Hi Nelle,
There is a small bug in the verbose option using python 3.7, with `--verbose 1`
```
Using iced version 0.5.6
Loading files...
Normalizing...
Filter 9 out of 237 bins ...
Matrix is t…
-
Most of the time the lccn data needs to be upper cased (hence a normalization in that direction in xmltold), but sometimes it doesn't... for instance
- `BQ7815 .K86 1990z` should have a final lower c…
eroux updated
5 years ago
-
_My previous comment was removed after reading a bit on the sRGB space._
We should add some documentation on the meaning of the scaled XYZ values and their relation to the sRGB representation that th…
-
For indexing purposes, it might be relevant to do some easy normalization of Sanskrit, mostly having *r+geminate* be normalized to *r+simple consonnant*. There are tons of examples in canonical collec…
eroux updated
3 years ago
-
I was going to comment at #745, but I think this translates to a more general discussion about vocabulary building. Although I don't know if this would be considered a meta issue.
## Character cove…
-
The following script composes a graphical model using `AnalysisFactor`'s and a `FactorGraphModel` (but fits them not using expectation propagation, but instead as just one big model):
https://gith…