Where communities thrive

  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
Repo info
    Aykut Çayır
    Hi guys, I have a problem covertype data set in mxnet mlp archtecture. When I start training with training data set, the process is not respobding or printing out. What could be the problem? Does anybody try this data set in mxnet mlp?
    Laurent Sorber
    Hi all, does anyone know if it is possible to have the nodes in a cluster train on different subsets of the data with XGBoost? Or does every node need to load in all of the data?
    Hi all, I have done some experimental work on Mesos integration on dmlc-core. I do not know whether or not do the dmlc guys like this.
    Punita Ojha
    This message was deleted
    Edward Z. Yang
    Hey folks! Is dmlc-core a header only library?
    Tianqi Chen
    it is we need compiled libs for the filesystems
    but many of the project depend on it only uses logging, json and serialization
    which are header only
    Rahul Bakshee
    hi guys . I am trying to implement a multi-class [0,1,2] classification problem. Can anyone tell me what should I initialize base_score with?

    I train my dataset on ubuntu system with python code.
    I test my model on MAC with c++ code.when I load the xxx.json,some errors happened:

    WorkSpace/mxnet/dmlc-core/include/dmlc/./json.h:842: JSONReader: Unknown field attrs, candidates are:

    @tqchen how to assign nnvm::Tuple<dmlc::optional<float>> a ? When I use a = nnvm::Tuple<float>{1.f, 2.f}; it seems wrong.
    Renato Marinho
    This message was deleted
    Renato Marinho
    This message was deleted
    Christian Clauss
    Anybody home? Would be interested to get reviews on https://github.com/dmlc/dmlc-core/pulls/cclauss
    hello anybody here?
    i see some implementation of std lib function like : 'any , ' in the dmlc-core source code
    since 'any' is in the standard lib from c++17 ,
    so is it true that the whole dmlc-core is designed for c++11 ?
    is this for the gpu programming cuda which only support c++11 ?
    can any one give some info about why the core is designed not using latest c++17 features?
    @tqchen @mli
    i checked that ' CUDA 10.0 supports C++14 but not C++17.'
    Vadim Dabravolski
    hey there, I'm trying to do multi-node/multi-GPU distributed training on AWS Sagemaker using one of GluonCV models. And i'd like to use DMLC tracker to run jobs using MPI protocol. Sagemaker MXNet container image comes with already pre-installed openmpi (specifically, "mpirun.real (OpenRTE) 4.0.1"). When i'm trying to mpi.submit I'm getting this exception: https://github.com/dmlc/dmlc-core/blob/dc0156a73a846751851fcf130729ade6360a37b1/tracker/dmlc_tracker/mpi.py#L35. Can someone confirm that mpirun.real is supported by DMLC? If yes, any examples on how to run the task on the cluster?
    Does anybody have an example of XGBoost in recommendation systems problem-solving (predict whether a given user would be interrested in an item/product)