Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support GPU training in the NVFlare demo #7965

Merged
merged 1 commit into from Jun 2, 2022
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
11 changes: 11 additions & 0 deletions demo/nvflare/README.md
Expand Up @@ -3,6 +3,8 @@
This directory contains a demo of Federated Learning using
[NVFlare](https://nvidia.github.io/NVFlare/).

## Training with CPU only

To run the demo, first build XGBoost with the federated learning plugin enabled (see the
[README](../../plugin/federated/README.md)).

Expand Down Expand Up @@ -53,3 +55,12 @@ Finally, shutdown everything from the admin CLI:
shutdown client
shutdown server
```

## Training with GPUs

To demo with Federated Learning using GPUs, make sure your machine has at least 2 GPUs.
Build XGBoost with the federated learning plugin enabled along with CUDA, but with NCCL
turned off (see the [README](../../plugin/federated/README.md)).

Modify `config/config_fed_client.json` and set `use_gpus` to `true`, then repeat the steps
above.
3 changes: 2 additions & 1 deletion demo/nvflare/config/config_fed_client.json
Expand Up @@ -12,7 +12,8 @@
"world_size": 2,
"server_cert_path": "server-cert.pem",
"client_key_path": "client-key.pem",
"client_cert_path": "client-cert.pem"
"client_cert_path": "client-cert.pem",
"use_gpus": "false"
}
}
}
Expand Down
7 changes: 6 additions & 1 deletion demo/nvflare/custom/trainer.py
Expand Up @@ -16,7 +16,7 @@ class SupportedTasks(object):

class XGBoostTrainer(Executor):
def __init__(self, server_address: str, world_size: int, server_cert_path: str,
client_key_path: str, client_cert_path: str):
client_key_path: str, client_cert_path: str, use_gpus: bool):
"""Trainer for federated XGBoost.

Args:
Expand All @@ -32,6 +32,7 @@ def __init__(self, server_address: str, world_size: int, server_cert_path: str,
self._server_cert_path = server_cert_path
self._client_key_path = client_key_path
self._client_cert_path = client_cert_path
self._use_gpus = use_gpus

def execute(self, task_name: str, shareable: Shareable, fl_ctx: FLContext,
abort_signal: Signal) -> Shareable:
Expand Down Expand Up @@ -66,6 +67,10 @@ def _do_training(self, fl_ctx: FLContext):

# Specify parameters via map, definition are same as c++ version
param = {'max_depth': 2, 'eta': 1, 'objective': 'binary:logistic'}
if self._use_gpus:
self.log_info(fl_ctx, f'Training with GPU {rank}')
param['tree_method'] = 'gpu_hist'
param['gpu_id'] = rank

# Specify validations set to watch performance
watchlist = [(dtest, 'eval'), (dtrain, 'train')]
Expand Down
8 changes: 7 additions & 1 deletion plugin/federated/README.md
Expand Up @@ -20,7 +20,12 @@ Build the Plugin
# Under xgboost source tree.
mkdir build
cd build
cmake .. -GNinja -DPLUGIN_FEDERATED=ON
# For now NCCL needs to be turned off.
cmake .. -GNinja\
-DPLUGIN_FEDERATED=ON\
-DUSE_CUDA=ON\
-DBUILD_WITH_CUDA_CUB=ON\
-DUSE_NCCL=OFF
ninja
cd ../python-package
pip install -e . # or equivalently python setup.py develop
Expand All @@ -31,5 +36,6 @@ Test Federated XGBoost
```shell
# Under xgboost source tree.
cd tests/distributed
# This tests both CPU training (`hist`) and GPU training (`gpu_hist`).
./runtests-federated.sh
```