From d7b67bb670dc1770a658304d98da149b7b29c37a Mon Sep 17 00:00:00 2001
From: Gokhan Ozbulak <gokhan.ozbulak@idiap.ch>
Date: Tue, 23 Apr 2024 22:09:42 +0200
Subject: [PATCH] File limit flag is added #60.
---
src/mednet/scripts/upload.py | 64 +++++++++++++++++++++++++-----------
1 file changed, 45 insertions(+), 19 deletions(-)
diff --git a/src/mednet/scripts/upload.py b/src/mednet/scripts/upload.py
index c8c7f8da..ec2ba7b2 100644
--- a/src/mednet/scripts/upload.py
+++ b/src/mednet/scripts/upload.py
@@ -89,6 +89,13 @@ def _create_temp_copy(source, target):
.. code:: sh
mednet upload --experiment-folder=/path/to/results --run-name=run-1
+
+4. Upload an existing experiment result with defining a size limit of 20MB for each file (set 0 for no limit):
+
+ .. code:: sh
+
+ mednet upload --experiment-folder=/path/to/results --file-size=20
+
""",
)
@click.option(
@@ -116,11 +123,22 @@ def _create_temp_copy(source, target):
help='A string indicating the run name (e.g. "run-1")',
cls=ResourceOption,
)
+@click.option(
+ "--file-limit",
+ "-l",
+ help='Limit file size to be uploaded in MB (set 0 for no limit).',
+ show_default=True,
+ required=True,
+ default=10,
+ type=click.IntRange(min=0),
+ cls=ResourceOption,
+)
@verbosity_option(logger=logger, cls=ResourceOption, expose_value=False)
def upload(
experiment_folder: pathlib.Path,
experiment_name: str,
run_name: str,
+ file_limit: int,
**_, # ignored
) -> None: # numpydoc ignore=PR01
"""Upload results from an experiment folder."""
@@ -146,7 +164,8 @@ def upload(
train_model_file, train_model_temp_file
)
with train_meta_file.open("r") as f:
- meta_data = json.load(f)
+ train_data = json.load(f)
+ train_files = [train_meta_file, train_log_file, train_model_temp_file]
# prepare evaluation files
evaluation_file = experiment_folder / "evaluation.json"
@@ -154,39 +173,46 @@ def upload(
evaluation_log_file = experiment_folder / "evaluation.pdf"
with evaluation_file.open("r") as f:
evaluation_data = json.load(f)
- test_data = evaluation_data["test"]
+ evaluation_data = evaluation_data["test"]
+ evaluation_files = [evaluation_file, evaluation_meta_file, evaluation_log_file]
+
+ # check for file sizes.
+ for f in train_files + evaluation_files:
+ file_size = f.stat().st_size / (1024**2)
+ if file_limit != 0 and file_size > file_limit:
+ raise RuntimeError(
+ f"Size of {f} ({file_size:.2f} MB) must be less than or equal to {file_limit} MB."
+ )
# prepare experiment and run names
experiment_name = (
experiment_name
if experiment_name
- else f'{meta_data["model-name"]}_{meta_data["database-name"]}'
+ else f'{train_data["model-name"]}_{train_data["database-name"]}'
)
- run_name = run_name if run_name else meta_data["datetime"]
+ run_name = run_name if run_name else train_data["datetime"]
logger.info("Setting experiment and run names on the MLFlow server...")
mlflow.set_experiment(experiment_name=experiment_name)
with mlflow.start_run(run_name=run_name):
# upload metrics
logger.info("Uploading metrics to MLFlow server...")
- mlflow.log_metric("threshold", test_data["threshold"])
- mlflow.log_metric("precision", test_data["precision"])
- mlflow.log_metric("recall", test_data["recall"])
- mlflow.log_metric("f1_score", test_data["f1_score"])
+ mlflow.log_metric("threshold", evaluation_data["threshold"])
+ mlflow.log_metric("precision", evaluation_data["precision"])
+ mlflow.log_metric("recall", evaluation_data["recall"])
+ mlflow.log_metric("f1_score", evaluation_data["f1_score"])
mlflow.log_metric(
- "average_precision_score", test_data["average_precision_score"]
+ "average_precision_score", evaluation_data["average_precision_score"]
)
- mlflow.log_metric("specificity", test_data["specificity"])
- mlflow.log_metric("auc_score", test_data["auc_score"])
- mlflow.log_metric("accuracy", test_data["accuracy"])
- mlflow.log_param("version", meta_data["package-version"])
+ mlflow.log_metric("specificity", evaluation_data["specificity"])
+ mlflow.log_metric("auc_score", evaluation_data["auc_score"])
+ mlflow.log_metric("accuracy", evaluation_data["accuracy"])
+ mlflow.log_param("version", train_data["package-version"])
# upload artifacts
logger.info("Uploading artifacts to MLFlow server...")
- mlflow.log_artifact(train_meta_file)
- mlflow.log_artifact(train_log_file)
- mlflow.log_artifact(train_model_temp_file)
- mlflow.log_artifact(evaluation_file)
- mlflow.log_artifact(evaluation_meta_file)
- mlflow.log_artifact(evaluation_log_file)
+ for f in train_files:
+ mlflow.log_artifact(f)
+ for f in evaluation_files:
+ mlflow.log_artifact(f)
# delete temporary file as no need it after logging.
train_model_temp_file.unlink()
--
GitLab