djl

djl copied to clipboard

Hello DJL team, I am trying to convert a network policy from Python using Pytorch to Java and one of the parameters I am missing in_channel and out_channel which is...

ranabameer

Error on training with multiple GPUs

3

## Copying an issue from [a Slack thread](https://deepjavalibrary.slack.com/archives/CPXNLGGTS/p1675956480760269): oleogin [1 month ago](https://deepjavalibrary.slack.com/archives/CPXNLGGTS/p1675956480760269) Hi I'm encountering an issue when trying to run training on multiple GPUs using DJL. I'm using the...

zachgk

Add pytorch engine support for percentile()

1

## Currently pytorch engine throws on percentile() or median(). Add native support for NDArray.percentile() and NDArray.median() This change will not change the API.

tklinchik

Running on Arm CPU

2

I've noticed that using DJL requires more than simply altering a Maven pom file and since I wish to do development on a machine that uses an Arm pricessor, I...

LobsterMan123

Warning: FALLBACK path has been taken inside: torch::jit::fuser::cuda::runCudaFusionGroup.

3

## Description Running the code specified below I get a number of warning at the beginning, apparently harmless. > Training: 0% |= | Accuracy: _, SoftmaxCrossEntropyLoss: _ > Training: 0%...

adepase

Linear class on loadMetadata method not initialize inputShapes

6

version: ai.djl api 0.21.0 ai.djl.nn.core.Linear#loadMetadata this method Override parent#loadMetadata cause inputShapes not initialize

luosoy

MXNet train error

2

VERSION: ai.djl.mxnet mxnet-engine 0.21.0 ai.djl.mxnet mxnet-native-mkl win-x86_64 1.9.1 Block: public static SequentialBlock build() { SequentialBlock resNet = new SequentialBlock(); resNet.add(Conv2d.builder() .setKernelShape(new Shape(3, 3)) // 9个图片数据合成1个数据 .setFilters(FILTERS_NUM) // 生成多少changle .optStride(new Shape(1,...

luosoy

Dimension mismatch or high dimensional dot operation is not supported

2

**Code** : NDManager manager = NDManager.newBaseManager(); NDArray X = manager.randomNormal(new Shape(1, 2)); NDArray trueW = manager.create(new float[]{2, -3.4f}); X.dot(trueW); **Exception** Dimension mismatch or high dimensional dot operation is not supported....

foolfish1

UnsatisfiedLinkError: C:\Users\kangert\AppData\Local\Temp\onnxruntime-java11604848578009472144\onnxruntime.dll: Can't find dependent libraries

3

## Description When linking the static library, I will be prompted that it cannot be found, but I have seen the temporary path created and it exists. What is the...

kkangert

djl runs failure with dynamic input format

3

## Description Use same onnx model as input to transform to tensorrt model with trtexec，if fix input format djl runs ok: ``` TensorRT-8.4.1.5/bin/trtexec --onnx=models/model.onnx --shapes=input_ids:1x4 --fp16 --saveEngine=model8415-1*4.trt ``` output: ```...

zjcDM