[BUG] Assertion Error while running xgboost example

[hwpang]

Describe the bug I am a new user to NVFlare. I tried running the example at ~/NVFlare/examples/advanced/xgboost/histogram-based, but I encountered the following error:

  File "/home/haowei/miniforge3/envs/nvflare_env/lib/python3.10/site-packages/nvflare/app_opt/xgboost/histogram_based/executor.py", line 277, in train
    with xgb.collective.CommunicatorContext(**communicator_env):
  File "/home/haowei/miniforge3/envs/nvflare_env/lib/python3.10/site-packages/xgboost/collective.py", line 280, in __enter__
    assert is_distributed()
AssertionError

To Reproduce Steps to reproduce the behavior:

1. Go to ~/NVFlare/examples/advanced/xgboost and run prepare_data.sh and prepare_job_config.sh
2. Go into histogram-based folder and run run_experiment_simulator.sh
3. See the same error trace

Expected behavior Expected to complete the example without error.

Desktop (please complete the following information):

• OS: Ubuntu 20.04
• Python Version: Python 3.10
• NVFlare Version: 2.4.1
• xgboost: 2.1.1

This seems to be related to https://github.com/dmlc/xgboost/pull/10503. However, note that I am using version of xgboost 2.1.1, which should contain the patch fix.

[YuanTingHsieh]

@hwpang hello, thanks for raising the issue.

Yes, the xgboost's repo's federated API has been changing in their newer releases. We have been working with them closely. We are working on updating our examples to accompany that.

Right now, please either use one of the following: (1) NVFlare main branch + xgboost from this: https://s3-us-west-2.amazonaws.com/xgboost-nightly-builds/list.html?prefix=federated-secure/ (2) NVFlare 2.4.0 with xgboost 2.0.3

[YuanTingHsieh]

Ok, after our fixes, please use one of the following: (1) NVFlare main branch + xgboost from this: https://s3-us-west-2.amazonaws.com/xgboost-nightly-builds/list.html?prefix=federated-secure/ (2) NVFlare 2.4 branch with xgboost 2.1.1

[hwpang]

Thanks, this works!