Minor fixes to xgboost example

[hwpang]

Description

Two minor fixes:

• Change the DATASET_ROOT environment variable in the notebook to $HOME/dataset. This is to be consistent with the path used in other scripts in this example, such as NVFlare/examples/advanced/xgboost/prepare_data.sh and NVFlare/examples/advanced/xgboost/histogram-based/run_experiment_centralized.sh.
• The training_mode argument should be training_algo:

    parser.add_argument(
        "--training_algo", type=str, default="bagging", choices=list(ALGO_DIR_MAP.keys()), help="Training algorithm"
    )

A few sentences describing the changes proposed in this pull request.

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Quick tests passed locally by running ./runtest.sh.
• In-line docstrings updated.
• Documentation updated.

[chesterxgchen]

@hwpang question: why prefer to use HOME directory to /data directory ? HOME directory is usually in the source code directory which something mixing the data and code together. /data at least refer the data in a different location, more close to the real case.

[hwpang]

@hwpang question: why prefer to use HOME directory to /data directory ? HOME directory is usually in the source code directory which something mixing the data and code together. /data at least refer the data in a different location, more close to the real case.

Thanks for the question. I was trying to make the path to the dataset consistent throughout this example. In the other scripts used in this example, the $HOME/dataset is assumed to be the directory of the dataset. For example, in NVFlare/examples/advanced/xgboost/prepare_data.sh, it looks for the dataset in $HOME/dataset:

DATASET_PATH="$HOME/dataset/HIGGS.csv"
OUTPUT_PATH="/tmp/nvflare/xgboost_higgs_dataset"
SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )

if [ ! -f "${DATASET_PATH}" ]
then
    echo "Please check if you saved HIGGS dataset in ${DATASET_PATH}"
fi

I don't have a strong preference of $HOME/dataset vs. /data, as long as the path is consistent with other scripts used in this example.

[ZiyueXu77]

@hwpang question: why prefer to use HOME directory to /data directory ? HOME directory is usually in the source code directory which something mixing the data and code together. /data at least refer the data in a different location, more close to the real case.

Thanks for the question. I was trying to make the path to the dataset consistent throughout this example. In the other scripts used in this example, the $HOME/dataset is assumed to be the directory of the dataset. For example, in NVFlare/examples/advanced/xgboost/prepare_data.sh, it looks for the dataset in $HOME/dataset:

DATASET_PATH="$HOME/dataset/HIGGS.csv"
OUTPUT_PATH="/tmp/nvflare/xgboost_higgs_dataset"
SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )

if [ ! -f "${DATASET_PATH}" ]
then
    echo "Please check if you saved HIGGS dataset in ${DATASET_PATH}"
fi

I don't have a strong preference of $HOME/dataset vs. /data, as long as the path is consistent with other scripts used in this example.

Thanks for noticing this! @chesterxgchen at the time we wanted to avoid creating new folders within examples, and avoid putting HIGGS to "/tmp" since it is big, so arbitrarily chose '$HOME'

[YuanTingHsieh]

I think consistency throughout notebook and the example folder is good, so this PR LGTM

[ZiyueXu77]

/build

[ZiyueXu77]

LGTM