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.

Sign up for GitHub

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

Jump to bottom

Add a gpu sample #575

Merged
merged 3 commits into from
Dec 21, 2018
Merged

Add a gpu sample #575

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
fcd6292
Add gpu sample
hongye-sun Dec 20, 2018
b31d6da
Merge branch 'master' into gpu-sample
hongye-sun Dec 20, 2018
c7db20d
Add gpu image to release script and update image version to commit hash
hongye-sun Dec 20, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions components/release.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ images=(
"ml-pipeline-dataflow-tfma"
"ml-pipeline-kubeflow-deployer"
"ml-pipeline-kubeflow-tf-trainer"
"ml-pipeline-kubeflow-tf-trainer-gpu"
"ml-pipeline-kubeflow-tf"
"ml-pipeline-dataproc-analyze"
"ml-pipeline-dataproc-create-cluster"
Expand Down
13 changes: 10 additions & 3 deletions samples/kubeflow-tf/kubeflow-training-classification.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,8 +35,8 @@ def dataflow_tf_transform_op(train_data: 'GcsUri', evaluation_data: 'GcsUri', sc
)


def kubeflow_tf_training_op(transformed_data_dir, schema: 'GcsUri[text/json]', learning_rate: float, hidden_layer_size: int, steps: int, target, preprocess_module: 'GcsUri[text/code/python]', training_output: 'GcsUri[Directory]', step_name='training'):
return dsl.ContainerOp(
def kubeflow_tf_training_op(transformed_data_dir, schema: 'GcsUri[text/json]', learning_rate: float, hidden_layer_size: int, steps: int, target, preprocess_module: 'GcsUri[text/code/python]', training_output: 'GcsUri[Directory]', step_name='training', use_gpu=False):
kubeflow_tf_training_op = dsl.ContainerOp(
name = step_name,
image = 'gcr.io/ml-pipeline/ml-pipeline-kubeflow-tf-trainer:85c6413a2e13da4b8f198aeac1abc2f3a74fe789',
arguments = [
Expand All @@ -51,6 +51,11 @@ def kubeflow_tf_training_op(transformed_data_dir, schema: 'GcsUri[text/json]', l
],
file_outputs = {'train': '/output.txt'}
)
if use_gpu:
kubeflow_tf_training_op.image = 'gcr.io/ml-pipeline/ml-pipeline-kubeflow-tf-trainer-gpu:85c6413a2e13da4b8f198aeac1abc2f3a74fe789',
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

.image needs to be an string, not tuple.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nice catch

kubeflow_tf_training_op.set_gpu_limit(1)

return kubeflow_tf_training_op

def dataflow_tf_predict_op(evaluation_data: 'GcsUri', schema: 'GcsUri[text/json]', target: str, model: 'TensorFlow model', predict_mode, project: 'GcpProject', prediction_output: 'GcsUri', step_name='prediction'):
return dsl.ContainerOp(
Expand Down Expand Up @@ -96,9 +101,11 @@ def kubeflow_training(output, project,
predict_mode='local'):
# TODO: use the argo job name as the workflow
workflow = '{{workflow.name}}'
# set the flag to use GPU trainer
use_gpu = True

preprocess = dataflow_tf_transform_op(train, evaluation, schema, project, preprocess_mode, '', '%s/%s/transformed' % (output, workflow)).apply(gcp.use_gcp_secret('user-gcp-sa'))
training = kubeflow_tf_training_op(preprocess.output, schema, learning_rate, hidden_layer_size, steps, target, '', '%s/%s/train' % (output, workflow)).apply(gcp.use_gcp_secret('user-gcp-sa'))
training = kubeflow_tf_training_op(preprocess.output, schema, learning_rate, hidden_layer_size, steps, target, '', '%s/%s/train' % (output, workflow), use_gpu=use_gpu).apply(gcp.use_gcp_secret('user-gcp-sa'))
prediction = dataflow_tf_predict_op(evaluation, schema, target, training.output, predict_mode, project, '%s/%s/predict' % (output, workflow)).apply(gcp.use_gcp_secret('user-gcp-sa'))
confusion_matrix = confusion_matrix_op(prediction.output, '%s/%s/confusionmatrix' % (output, workflow)).apply(gcp.use_gcp_secret('user-gcp-sa'))

Expand Down