Make s3 bucket name and s3 region required parameters in dashboard wr…

…apper. Make hardware_metrics optional. Fix custom csv path issue when fetching from s3 (#46) * Make s3 bucket name and s3 region required parameters in dashboard wrapper. Make hardware_metrics optional. Fix custom csv path issue when fetching from s3 * remove DEFAULT in csv path * formatting * change csv defaults * ignore csv paths file + remove print statement * address PR comments * fix unittest * ignore csv_paths file to prevent changing of defaults
autogluon · Aug 10, 2023 · 36b4511 · 36b4511
1 parent 7256999
commit 36b4511
Show file tree

Hide file tree

Showing 5 changed files with 36 additions and 51 deletions.
diff --git a/.gitignore b/.gitignore
@@ -6,6 +6,7 @@ web_files
 *.csv
 coverage.json
 results.xml
+csv_paths.py
 
 #Venv
 .venv*/

diff --git a/src/autogluon_dashboard/app.py b/src/autogluon_dashboard/app.py
@@ -16,10 +16,10 @@
     PER_DATA_COMP,
     PER_DATASET_IDF,
 )
-from autogluon_dashboard.constants.aws_s3_constants import (
-    AGG_FRAMEWORK_DEFAULT_CSV_PATH,
-    HARDWARE_METRICS_DEFAULT_CSV_PATH,
-    PER_DATASET_DEFAULT_CSV_PATH,
+from autogluon_dashboard.constants.csv_paths import (
+    AGG_FRAMEWORK_CSV_PATH,
+    HARDWARE_METRICS_CSV_PATH,
+    PER_DATASET_CSV_PATH,
 )
 from autogluon_dashboard.constants.df_constants import (
     BESTDIFF,
@@ -77,9 +77,9 @@
 from autogluon_dashboard.widgets.slider_widget import SliderWidget
 
 # Load Data
-dataset_file = os.environ.get("PER_DATASET_S3_PATH", PER_DATASET_DEFAULT_CSV_PATH)
-aggregated_file = os.environ.get("AGG_DATASET_S3_PATH", AGG_FRAMEWORK_DEFAULT_CSV_PATH)
-hardware_metrics_file = os.environ.get("HWARE_METRICS_S3_PATH", HARDWARE_METRICS_DEFAULT_CSV_PATH)
+dataset_file = PER_DATASET_CSV_PATH
+aggregated_file = AGG_FRAMEWORK_CSV_PATH
+hardware_metrics_file = HARDWARE_METRICS_CSV_PATH
 dataset_paths = [dataset_file, aggregated_file, hardware_metrics_file]
 per_dataset_df, all_framework_df, hware_metrics_df = get_dataframes(dataset_paths)
 
@@ -122,9 +122,10 @@
 per_dataset_csv_widget = FileDownloadWidget(file=PER_DATASET_DOWNLOAD_TITLE).create_widget()
 all_framework_df.to_csv(AGG_FRAMEWORKS_DOWNLOAD_TITLE)
 all_framework_csv_widget = FileDownloadWidget(file=AGG_FRAMEWORKS_DOWNLOAD_TITLE).create_widget()
+hware_metrics_csv_widget = None
 if hware_metrics_df is not None:
     hware_metrics_df.to_csv(HARDWARE_METRICS_DOWNLOAD_TITLE)
-hware_metrics_csv_widget = FileDownloadWidget(file=HARDWARE_METRICS_DOWNLOAD_TITLE).create_widget()
+    hware_metrics_csv_widget = FileDownloadWidget(file=HARDWARE_METRICS_DOWNLOAD_TITLE).create_widget()
 
 df_framework_only = get_df_filter_by_framework(per_dataset_idf, frameworks_widget3)
 prop_best = get_proportion_framework_rank1(df_framework_only, per_dataset_df, len(dataset_list))

diff --git a/src/autogluon_dashboard/constants/aws_s3_constants.py b/src/autogluon_dashboard/constants/aws_s3_constants.py
@@ -1,13 +1,2 @@
-S3_REGION = "us-east-2"
-DEFAULT_BUCKET_NAME = "ag-dashboard-test"  # TODO: Change default bucket name
 CSV_FILES_DIR = "dev_data/"
-PER_DATASET_DEFAULT_CSV_PATH = (
-    "https://d24iwcyhf6yavw.cloudfront.net/dev_data/all_data.csv"  # TODO: Change default csv path
-)
-AGG_FRAMEWORK_DEFAULT_CSV_PATH = (
-    "https://d24iwcyhf6yavw.cloudfront.net/dev_data/autogluon.csv"  # TODO: Change default csv path
-)
-HARDWARE_METRICS_DEFAULT_CSV_PATH = (
-    "https://d24iwcyhf6yavw.cloudfront.net/dev_data/hardware_metrics.csv"  # TODO: Change default csv path
-)
 CLOUDFRONT_DOMAIN = "https://d24iwcyhf6yavw.cloudfront.net"
diff --git a/src/autogluon_dashboard/constants/csv_paths.py b/src/autogluon_dashboard/constants/csv_paths.py
@@ -0,0 +1,3 @@
+PER_DATASET_CSV_PATH = "https://d24iwcyhf6yavw.cloudfront.net/dev_data/all_data.csv"
+AGG_FRAMEWORK_CSV_PATH = "https://d24iwcyhf6yavw.cloudfront.net/dev_data/autogluon.csv"
+HARDWARE_METRICS_CSV_PATH = "https://d24iwcyhf6yavw.cloudfront.net/dev_data/hardware_metrics.csv"
diff --git a/src/autogluon_dashboard/dashboard.py b/src/autogluon_dashboard/dashboard.py
@@ -6,12 +6,7 @@
 import boto3
 import botocore
 
-from autogluon_dashboard.constants.aws_s3_constants import (
-    CLOUDFRONT_DOMAIN,
-    CSV_FILES_DIR,
-    DEFAULT_BUCKET_NAME,
-    S3_REGION,
-)
+from autogluon_dashboard.constants.aws_s3_constants import CLOUDFRONT_DOMAIN, CSV_FILES_DIR
 
 
 def upload_to_s3(s3_client: botocore.client, file_name: str, object_name: str, bucket_name: str, args: dict = None):
@@ -51,14 +46,15 @@ def get_args() -> argparse.Namespace:
     parser.add_argument(
         "--hware_metrics_csv",
         type=str,
-        required=True,
         help="Location of csv file of hardware metrics in local filesystem to upload to S3 bucket. Example: sub_folder/file_name.csv",
+        default="",
         metavar="",
     )
 
     parser.add_argument(
         "--s3_bucket",
         type=str,
+        required=True,
         help="Name of S3 bucket that results to aggregate get outputted to",
         nargs="?",
         metavar="",
@@ -74,6 +70,7 @@ def get_args() -> argparse.Namespace:
     parser.add_argument(
         "--s3_region",
         type=str,
+        required=True,
         help="S3 Region to deploy the dashboard website",
         nargs="?",
         metavar="",
@@ -85,7 +82,7 @@ def get_args() -> argparse.Namespace:
 
 def run_dashboard():
     args = get_args()
-    # Set variables to corrensponding command line args
+    # Set variables to corresponding command line args
     per_dataset_csv_path = args.per_dataset_csv
     aggregated_csv_path = args.agg_dataset_csv
     hware_metrics_csv_path = args.hware_metrics_csv
@@ -96,24 +93,7 @@ def run_dashboard():
     logger = logging.getLogger("dashboard-logger")
     logging.basicConfig(level=logging.INFO)
 
-    if not bucket_name:
-        if region:
-            logger.warning(
-                "Cannot specify region if no bucket has been provded. Defaulting to AutoGluon bucket and region (%s)",
-                S3_REGION,
-            )
-        else:
-            logger.info(
-                "No bucket or region has been provided. Defaulting to AutoGluon bucket and region (%s)",
-                S3_REGION,
-            )
-        region = S3_REGION
-        bucket_name = DEFAULT_BUCKET_NAME  # TODO: Change default bucket name
-    else:
-        if not region:
-            raise ValueError("You must specify a region if you provide a bucket")
-        else:
-            logger.info(f"You have specified Bucket {bucket_name} and Region {region}.")
+    logger.info(f"You have specified Bucket {bucket_name} and Region {region}.")
 
     # Set S3 URL with appropriate bucket, region, and prefix
     if prefix:
@@ -127,22 +107,32 @@ def run_dashboard():
     per_dataset_s3_loc = CSV_FILES_DIR + "all_data.csv"
     aggregated_s3_loc = CSV_FILES_DIR + "autogluon.csv"
     hware_s3_loc = CSV_FILES_DIR + "hardware_metrics.csv"
-    os.environ["PER_DATASET_S3_PATH"] = CLOUDFRONT_DOMAIN + f"/{prefix}" + per_dataset_s3_loc
-    os.environ["AGG_DATASET_S3_PATH"] = CLOUDFRONT_DOMAIN + f"/{prefix}" + aggregated_s3_loc
-    os.environ["HWARE_METRICS_S3_PATH"] = CLOUDFRONT_DOMAIN + f"/{prefix}" + hware_s3_loc
+    PER_DATASET_CSV_PATH = CLOUDFRONT_DOMAIN + f"/{prefix}" + per_dataset_s3_loc
+    AGG_FRAMEWORK_CSV_PATH = CLOUDFRONT_DOMAIN + f"/{prefix}" + aggregated_s3_loc
+    HARDWARE_METRICS_CSV_PATH = CLOUDFRONT_DOMAIN + f"/{prefix}" + hware_s3_loc if hware_metrics_csv_path else ""
+    wrapper_dir = os.path.dirname(__file__)
+    csv_path_file_location = os.path.join(wrapper_dir, "constants/csv_paths.py")
+    # Write the CSV paths to a python file that can be accessed by app.py
+    # We do this instead of using an environment variable since the website runs in a separate web environment, which cannot be accessed preemptively using this code.
+    f = open(csv_path_file_location, "w")
+    f.write(f"PER_DATASET_CSV_PATH = " + '"' + PER_DATASET_CSV_PATH + '"')
+    f.write(f"\nAGG_FRAMEWORK_CSV_PATH = " + '"' + AGG_FRAMEWORK_CSV_PATH + '"')
+    f.write(f"\nHARDWARE_METRICS_CSV_PATH = " + '"' + HARDWARE_METRICS_CSV_PATH + '"')
+    f.close()
+
     os.environ["BOKEH_PY_LOG_LEVEL"] = "error"
 
     s3_client = boto3.client("s3")
 
     # Upload CSV files to S3
     upload_to_s3(s3_client, per_dataset_csv_path, prefix + per_dataset_s3_loc, bucket_name)
     upload_to_s3(s3_client, aggregated_csv_path, prefix + aggregated_s3_loc, bucket_name)
-    upload_to_s3(s3_client, hware_metrics_csv_path, prefix + hware_s3_loc, bucket_name)
+    if hware_metrics_csv_path:
+        upload_to_s3(s3_client, hware_metrics_csv_path, prefix + hware_s3_loc, bucket_name)
     logger.info(
         f"Evaluation CSV files have been successfully uploaded to bucket - {bucket_name}, at locations: {s3_url + per_dataset_s3_loc}, {s3_url + aggregated_s3_loc}, and {s3_url + hware_s3_loc}.",
     )
 
-    wrapper_dir = os.path.dirname(__file__)
     agg_script_location = os.path.join(wrapper_dir, "utils/aggregate_file.py")
     agg_file_location = os.path.join(wrapper_dir, "out.py")
     # Aggregate all code into output file
@@ -176,8 +166,9 @@ def run_dashboard():
     upload_to_s3(s3_client, os.path.join(web_files_dir, "out.js"), prefix + "out.js", bucket_name)
     logger.info("WebAssembly files have been successfully uploaded to bucket - %s", bucket_name)
 
-    # TODO: Change website link to https using CloudFront
     logger.info("The dashboard website is: " + f"{CLOUDFRONT_DOMAIN}/{prefix}out.html")
+    # Use print so that the GitHub Actions bash script can pick up the URL from the CLI
+    print("The dashboard website is: " + f"{CLOUDFRONT_DOMAIN}/{prefix}out.html")
 
 
 if __name__ == "__main__":