IndexOutOfBoundsException while importing a model from configuration file

[axyonovleonid]

Hi,

Kotlin DL v. 0.0.2 color_model.txt

I'm trying to adapt this (see 'Build & Compile the Model') section network as it shown at importing existing nn guide. I've exported model config in json format (config gist or color_model.txt in attachments) and when I try call Sequential.loadModelConfiguration(config) I get an exeption

Caused by: java.lang.IndexOutOfBoundsException: Index 2 out of bounds for length 2 at java.base/jdk.internal.util.Preconditions.outOfBounds(Preconditions.java:64) at java.base/jdk.internal.util.Preconditions.outOfBoundsCheckIndex(Preconditions.java:70) at java.base/jdk.internal.util.Preconditions.checkIndex(Preconditions.java:248) at java.base/java.util.Objects.checkIndex(Objects.java:373) at java.base/java.util.ArrayList.get(ArrayList.java:425) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.loadSequentialModelLayers(ModelLoader.kt:115) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.loadModelConfiguration(ModelLoader.kt:44) at org.jetbrains.kotlinx.dl.api.core.Sequential$Companion.loadModelConfiguration(Sequential.kt:66) at com.my.nets.ColorClassifier.<init>(ColorClassifier.kt:17) at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method) at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62) at java.base/jdk.internal.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)

I've figured out that exception occurs at "batch_input_shape" array independently from it's size and content. Am I doing smth wrong or there really is a problem with reading?

[zaleslaw]

Could you please try the same thing with the new version 0.3.0, which will be released today? Probably this problem will be fixed? Thanks for providing so much helpful feedback on this issue. I'll try to reproduce and return in 2-3 days.

Meanwhile, you could try a new version and share results here in the issue.

[axyonovleonid]

Yeah, it's gone in the new version. But another problems occured. Now I get NLP at model configuration stage

Caused by: java.lang.NullPointerException: null at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.convertToRegularizer(ModelLoader.kt:225) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.createDenseLayer(ModelLoader.kt:557) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.convertToLayer(ModelLoader.kt:77) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.loadSequentialModelLayers(ModelLoader.kt:65) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.deserializeSequentialModel(ModelLoader.kt:45) at org.jetbrains.kotlinx.dl.api.inference.keras.ModelLoaderKt.loadSequentialModelConfiguration(ModelLoader.kt:41) at org.jetbrains.kotlinx.dl.api.core.Sequential$Companion.loadModelConfiguration(Sequential.kt:69) at com.my.nets.ColorClassifier.<init>(ColorClassifier.kt:17) at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method) at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62) at java.base/jdk.internal.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45) at java.base/java.lang.reflect.Constructor.newInstanceWithCaller(Constructor.java:500) at java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:481) at kotlin.reflect.jvm.internal.calls.CallerImpl$Constructor.call(CallerImpl.kt:41) at kotlin.reflect.jvm.internal.KCallableImpl.call(KCallableImpl.kt:108) at kotlin.reflect.jvm.internal.KCallableImpl.callDefaultMethod$kotlin_reflection(KCallableImpl.kt:159) at kotlin.reflect.jvm.internal.KCallableImpl.callBy(KCallableImpl.kt:112) at org.springframework.beans.BeanUtils$KotlinDelegate.instantiateClass(BeanUtils.java:798) at org.springframework.beans.BeanUtils.instantiateClass(BeanUtils.java:189) ... 22 common frames omitted

I tried to change NN config in Jupyter and specify all of parameters I see in model.json and replace other null fields in config with zeros or smth valueable, but it didn't help. Actually I'm quite new to NN projecting, so I'm not sure it should have helped

Models model_as_is_from_jupyter.txt

model_with_null_avoidance.txt