-
#### Issue Description
Currently, for ComputationGraph and unlike what exists for MultiLayerNetwork, there is no output(INDArray[] inputs, boolean train, INDArray[] featuresMasks, INDArray[] labels…
-
#### Issue Description
SequenceRecordReaderDataSetIterator calls preProcess twice per next()
It's called already at the ed of mdsToDataSet() and directly after it returns again:
-----------------…
-
Using the current head of the master, several unit-tests fail. This makes it really hard to properly prepare PRs for the repo since it hard to track what test failures are present already, and which…
-
#### Issue Description
During training at the start of each epoch (pass through data) the UI is showing me that the log10(stdDev) of the inputs is going up by 1.5 orders of magnitude.
See the atta…
-
#### Issue Description
There is an odd behaviour when loading data using `SequenceRecordReaderDataSetIterator` with a normalizer. It seems like the normalizer is applied twice on returned `DataSet`…
ghost updated
5 years ago
-
I have three .csv files: `words.csv`, `feats.csv`, `labels.csv` with the following content.
words.csv contains vocab indeces
```
2134
213212
112
335
224949
```
feats.csv contains correspon…
-
Is there any way to use the column key and title like this
` {title: "Name", dataKey: "name"},` for multidataset.
-
I am able to download an excel file with the example provided, i.e. Excel Export with custom cells style. However, the styles could not be applied to the cells. Am I missing out on something for the e…
-
When I change export OMP_NUM_THREADS=10, I still see that num of threads is one!
> 03:43:34.520 [main] DEBUG org.nd4j.nativeblas.NativeOps - Number of threads used for linear algebra 1
> 03:43:34.530…
-
My goal is to predict the label of a set of sentences.
Therefore, the input for my problem is a set of sentences, masked and inserted to an LSTM (w/ shared parameters using StackVertex and UnstackVer…