mxnet

mxnet copied to clipboard

[FEATURE] Add query_keys transformer version without split

2

## Description ## MXNet is fusing split, reshape, swapaxis and batch_dot operators for performance purpose. In gpt-2 model this fuse could be done as well if we exclude split. ...

agrabows

[BUGFIX] _npi_repeats with swap

15

## Description ## _npi_repeats gave correct outputs only when repetition happened on 0th axis. That's why swapaxes function was called before and after triggering _npi_repeats operator. swapaxes() was called from...

Kacper-Pietkun

Issue when running Distributed Training with Sparse Gradients

1

## Description (A clear and concise description of what the bug is.) When running distributed training (multi-instance with each instance having a single GPU) with sparse gradients (produced by negative...

BlakeLazarine

Add support for bool data type for condition in where operator

5

## Description ## This change allows to utilize bool data type for condition tensor in oneDNN implementation of where operator ## Checklist ## ### Essentials ### - [ ] PR's...

bgawrych

cuDNN batchnorm behaviour is not consistent and it can output nan

4

This script creates a batchnorm and runs it 3 times: 1) A first test mode evaluation 2) A dummy training mode evaluation 3) A second test mode evaluation The outputs...

matteosal

Building errors after cmake

2

## Description Trying to build mxnet for Ubuntu as explained here: https://mxnet.apache.org/versions/1.9.1/get_started/ubuntu_setup.html I get the following errors (_Fehler_ is the german word for _error_) ### Error Message /home/gigi/Downloads/mxnet/src/operator/contrib/bilinear_resize-inl.cuh(228): warning: variable...

wilhem

Cannot use pip to install mxnet via VScode

2

## Description Cannot use pip to install mxnet via VScode: Every time I use the command 'pip install mxnet', it always stuck at here for a while: '``` Using legacy...

Jason-will3

Flaky CI step1- cannot clear workspace directory

1

## Description I'm seeing CI jobs fail at the first step `Recursively delete the current directory from the workspace`. Log output is: ``` java.nio.channels.ClosedChannelException ``` Seems unrelated to the specific...

DickJC123

CI Node nvml configuration issue for website and centos-gpu pipelines

1

## Description log output: ``` nvidia-container-cli: initialization error: nvml error: driver/library version mismatch: unknown. ``` This can be caused by a mismatch between the installed driver and the currently loaded...

DickJC123

support cuda version higher than 11.2

3

Hi, It seems like that newest version is mxnet-cu112. Will there be mxnet-cu114 or support higher version? Thanks!

garynlfd