Skip to content

Commit

Permalink
Update test image and close file (#19)
Browse files Browse the repository at this point in the history
  • Loading branch information
adheshgarg authored Sep 10, 2024
1 parent f5c8034 commit bf7adc6
Show file tree
Hide file tree
Showing 3 changed files with 7 additions and 7 deletions.
4 changes: 2 additions & 2 deletions examples/basic-job-example-config.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ hydra:

training_cfg:
# entry_script: Required. Path to the entry script of training/fine-tuning, this path should be inside container
entry_script: ./train.py
entry_script: /opt/pytorch-mnist/mnist.py
# script_args: Optional. List of script arguments. Example of usage:
# script_args:
# - --max_context_width: 4096
Expand Down Expand Up @@ -90,7 +90,7 @@ cluster:
# base_results_dir: Optional. Location to store the results, checkpoints and logs.
base_results_dir: ./result
# container: Required. Docker image to be used for Training Job
container: docker.io/kubeflowkatib/pytorch-mnist:v1beta1-45c5727
container: docker.io/kubeflowkatib/pytorch-mnist-cpu:v1beta1-bc09cfd

# env_vars: Optional. Environment variables passed to the training job.
env_vars:
Expand Down
6 changes: 3 additions & 3 deletions src/hyperpod_cli/utils.py
Original file line number Diff line number Diff line change
Expand Up @@ -105,11 +105,11 @@ def store_current_hyperpod_context(data):


def _retrieve_current_hyperpod_context():
file = open(
with open(
GENERATED_LAUNCHER_CONFIG_FILE_PATH + HYPERPOD_CLUSTER_CONTEXT_FILE_NAME,
"r",
)
return json.load(file)
) as file:
return json.load(file)


def _validate_link(console_url):
Expand Down
4 changes: 2 additions & 2 deletions test/integration_tests/data/basicJob.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ hydra:
output_subdir: null

training_cfg:
entry_script: ./train.py
entry_script: /opt/pytorch-mnist/mnist.py
script_args: []
run:
name: hyperpod-cli-test # Current run name
Expand Down Expand Up @@ -49,7 +49,7 @@ cluster:
restartPolicy: OnFailure # restart policy

base_results_dir: ./result # Location to store the results, checkpoints and logs.
container: docker.io/kubeflowkatib/pytorch-mnist:v1beta1-45c5727 # container to use
container: docker.io/kubeflowkatib/pytorch-mnist-cpu:v1beta1-bc09cfd # container to use

env_vars:
NCCL_DEBUG: INFO # Logging level for NCCL. Set to "INFO" for debug information

0 comments on commit bf7adc6

Please sign in to comment.