udacity/udacity-nd009t-c2-developing-ml-workflow

udacity-nd009t-c2-developing-ml-workflow's Issues

lesson4/exercises-solutions.ipynb, sampling_percentage=100 (course content says =34)

The course exercise calls for setting sampling_percentage=34, but this solution code for that exercise sets it to 100.

# SOLUTION
from sagemaker.model_monitor import DataCaptureConfig

capture_uri = f's3://{bucket}/data-capture'

data_capture_config = DataCaptureConfig(
    enable_capture=True,
    sampling_percentage=100,
    destination_s3_uri=capture_uri
)

project/starter.ipynb function generate_test_case() not working as it currently exists.

The following function, as it exists in the repo, would not work for me or for others, until changing this:
objects = s3.Bucket(bucket).objects.filter("test")
to this:
objects = s3.Bucket(bucket).objects.filter(Prefix="test")

def generate_test_case():
    # Setup s3 in boto3
    s3 = boto3.resource('s3')
    
    # Randomly pick from sfn or test folders in our bucket
    objects = s3.Bucket(bucket).objects.filter("test")
    
    # Grab any random object key from that folder!
    obj = random.choice([x.key for x in objects])
    
    return json.dumps({
        "image_data": "",
        "s3_bucket": bucket,
        "s3_key": obj
    })
generate_test_case()

Add more details on model training section of the project

It would be a lot easier to know on average how long usually takes to train the model.

Motivation for the change:
As shown in the screenshot below, training is only at epoch 2 of 30 which took over 800seconds already.