Alex Black

Results 38 issues of Alex Black

Trying to load a RampSchedule in a ComputationGraph: ``` Caused by: org.nd4j.shade.jackson.databind.exc.InvalidDefinitionException: Cannot construct instance of `org.nd4j.linalg.schedule.RampSchedule` (no Creators, like default construct, exist): cannot deserialize from Object value (no delegate-...

Bug
DL4J

Not a priority for us, but this might be neat to add: https://github.com/zalandoresearch/fashion-mnist The main thing is that it's a drop-in replacement (same data format) as MNIST/EMNIST, so it should...

Enhancement
DL4J
good first issue

This works fine: ``` @Test public void test5(){ INDArray arr = Nd4j.linspace(DataType.FLOAT, 1, 6, 1).reshape(2,3); INDArray out = Nd4j.exec(new Reverse(arr))[0]; System.out.println(out); } ``` ``` [[ 6.0000, 5.0000, 4.0000], [ 3.0000,...

ND4J

Used only in DL4J UI. To avoid version incompatibilities like this: https://github.com/eclipse/deeplearning4j/issues/8725

GUI

``` @Test public void testUniformDtype2(){ for( int i=0; i

Bug
SameDiff

**For bugs, open a [GitHub issue in the eclipse/deeplearning4j repository](https://github.com/eclipse/deeplearning4j/issues)** **For general support, please use Discourse: https://community.konduit.ai** Welcome to the Github issues section for the Eclipse Deeplearning4j, ND4J, RL4J, SameDiff,...

Idea: If vectors are not detected/available, give user the option to download to a default location. The user should be warned about the size and have to confirm the intention...

Network seems to learn well; then score increases at the end, becoming NaNs in some runs. Probably tuning/regularization related. Tested on DL4J master. https://gist.github.com/AlexDBlack/37bc528fd40e77cde0586801e6f53cce

For a Rust error type with associated data called `TestError`, the Kotlin bindings generator generates a `sealed class TestError`, yet functions throwing that error try to call it via `rustCallWithError(TestException)`....

Related: https://github.com/mozilla/uniffi-rs/issues/1062 Note it's a different problem and fix required (i.e., fixing #1062 won't fix this), but this Swift code generation issue can be reproduced with the same test case....