fix: add in MODEL_WEIGHTS, MODEL_LABELS to cdk stack

README.md

@@ -4,9 +4,12 @@
 [![Python](https://img.shields.io/badge/language-Python-blue.svg)](https://www.python.org/downloads/)
 
 **fastapi-yolov5** code deploys and runs the YOLOv5 model in the Python web framework [FastAPI](https://fastapi.tiangolo.com/) either locally or in AWS.
+In AWS, the model is deployed with ECS Fargate and exposed with an Application Load Balancer.
 
 It is currently live at https://deepsea-ai.mbari.org/megadetector/docs
 
+![Image link ](api_example.png)
+
 # Requirements
 
 - [Docker](https://docs.docker.com/get-docker/)
@@ -43,15 +46,15 @@ docker-compose down
 # Running
 
 ## Health Check
-Check the health of the server by going to `http://localhost:3000/health`.  You should see the following response:
+Check the health of the server by going to `http://localhost:8000/health`.  You should see the following response:
 
 ```json
 {"status":"ok"}
 ```
 
 ## Predict to JSON
 
-Send a POST request to `http://localhost:3000/predict` with an image file in the body to get a prediction returned in JSON format.
+Send a POST request to `http://localhost:8000/predict` with an image file in the body to get a prediction returned in JSON format.
 By default, predictions greater than 0.01 are posted.
 
 ```shell
@@ -120,7 +123,7 @@ app:
     environment:
       - MODEL_WEIGHTS=/app/models/best/best.pt
       - MODEL_LABELS=/app/models/best/labels.txt
-      - MODEL_DESCRIPTION="Megadetector"
+      - MODEL_DESCRIPTION=Megadetector
 ```
  
 The labels file should be in the format of one label per line.
@@ -141,16 +144,22 @@ docker-compose up
 
 ## Deploying a custom model in AWS
 
-To override the default model, upload the `best.pt` file and the labels file `labels.txt` for that model to an S3 bucket.
-Specify the S3 bucket in the `MODELPATH` environment variable in the `config.yaml` file.
-The S3 bucket must be in the same region as the ECS cluster
+Specify any custom configuration on scaling in `config.yaml` file for the stack resources, e.g. min and max capacity.
 
 ```yaml
-app:
-    environment:
-      - MODEL_WEIGHTS=s3://901103-models-deploy/megadetector/best.pt
-      - MODEL_LABELS=s3://901103-models-deploy/megadetector/best.pt 
-      - MODEL_DESCRIPTION="Megadetector"
+MinCapacity: 1
+MaxCapacity: 5
+```
+
+To override the default model, upload yor model, e.g. `best.pt` file and the labels file `labels.txt` for that model to an S3 bucket.
+The S3 bucket must be in the same region as the ECS cluster. Define the S3 bucket and the model path in the `config.yml` file.
+These can be overridden as needed in the AWS console through the ECS task definition as environment variables.
+
+```yaml
+MODEL_WEIGHTS: s3://901103-models-deploy/midwatervars102/best.pt
+MODEL_LABELS: s3://901103-models-deploy/midwatervars102/labels.txt
+MODEL_DESCRIPTION: Megadetector
+MODEL_INPUT_SIZE: 1280
 ```
 
 Deploy the stack with the new configuration

cdk/fastapi.py

@@ -40,12 +40,13 @@ def __init__(
         with open("config.yml", 'r') as stream:
             config = yaml.safe_load(stream)
 
-        # Get the model path, and capacity from the config file
-        model_path = config['ModelPath']
+        # Get the model information from environment variables
+        model_weights = config['ModelWeights']
+        model_labels = config['ModelLabels']
+
+        # Cluster capacity
         min_capacity = config['MinCapacity']
         max_capacity = config['MaxCapacity']
-        model_description = config['ModelDescription']
-        model_input_size = config['ModelInputSize']
 
         # Create VPC
         vpc = ec2.Vpc(self, "FastAPIYOLOVv5VPC", max_azs=2)
@@ -74,8 +75,6 @@ def __init__(
             connection=ec2.Port.tcp(80)
         )
 
-        # Retrieve the secret value from AWS Secrets Manager
-
         # Retrieve the AWS access key ID secret value from AWS Secrets Manager
         secret = secrets_manager.Secret.from_secret_name_v2(self, "MySecretID", secret_name="prod/s3download")
 
@@ -87,9 +86,10 @@ def __init__(
             image=docker_image,
             container_port=80,
             environment={
-                "MODEL_PATH": model_path,
-                "MODEL_DESCRIPTION": model_description,
-                "MODEL_INPUT_SIZE": str(model_input_size),
+                "MODEL_WEIGHTS": config['MODEL_WEIGHTS'],
+                "MODEL_LABELS": config['MODEL_LABELS'],
+                "MODEL_DESCRIPTION": config['MODEL_DESCRIPTION'],
+                "MODEL_INPUT_SIZE": str(config['MODEL_INPUT_SIZE']),
             },
             secrets={
                 "AWS_ACCESS_KEY_ID": ecs.Secret.from_secrets_manager(secret, "AWS_ACCESS_KEY_ID"),

config.yml

@@ -7,7 +7,7 @@ Account: 975513124282
 ProjectName: FastAPI-YOLOv5
 ProjectNumber: 901103-yv5
 Description: This stack deploys a Fargate managed ELB environment for creating localized detections from a YOLOv5 model using the FastAPI framework with a general underwater megadetector
-DockerURI: public.ecr.aws/e5z9d1o0/mbari/pypampsd:444bb3b486087df80ff38bdf05683ec52fd180c1
-ModelPath: s3://901103-models-deploy/megadetector/
-ModelDescription: Megadetector
-ModelInputSize: 1280
+MODEL_WEIGHTS: s3://901103-models-deploy/midwatervars102/best.pt
+MODEL_LABELS: s3://901103-models-deploy/midwatervars102/labels.txt
+MODEL_DESCRIPTION: Megadetector
+MODEL_INPUT_SIZE: 1280